dm.c 65.4 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
 *
 * This file is released under the GPL.
 */

8 9
#include "dm-core.h"
#include "dm-rq.h"
M
Mike Anderson 已提交
10
#include "dm-uevent.h"
L
Linus Torvalds 已提交
11 12 13

#include <linux/init.h>
#include <linux/module.h>
A
Arjan van de Ven 已提交
14
#include <linux/mutex.h>
15
#include <linux/sched/signal.h>
L
Linus Torvalds 已提交
16 17 18
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/mempool.h>
19
#include <linux/dax.h>
L
Linus Torvalds 已提交
20 21
#include <linux/slab.h>
#include <linux/idr.h>
D
Darrick J. Wong 已提交
22
#include <linux/hdreg.h>
23
#include <linux/delay.h>
24
#include <linux/wait.h>
25
#include <linux/pr.h>
26

27 28
#define DM_MSG_PREFIX "core"

N
Namhyung Kim 已提交
29 30 31 32 33 34 35 36 37 38
#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
39 40 41 42 43 44 45
/*
 * 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
46 47 48 49 50
static const char *_name = DM_NAME;

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

51 52
static DEFINE_IDR(_minor_idr);

53
static DEFINE_SPINLOCK(_minor_lock);
M
Mikulas Patocka 已提交
54 55 56 57 58

static void do_deferred_remove(struct work_struct *w);

static DECLARE_WORK(deferred_remove_work, do_deferred_remove);

59 60
static struct workqueue_struct *deferred_remove_workqueue;

61 62 63
atomic_t dm_global_event_nr = ATOMIC_INIT(0);
DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq);

L
Linus Torvalds 已提交
64 65 66 67 68
/*
 * One of these is allocated per bio.
 */
struct dm_io {
	struct mapped_device *md;
69
	blk_status_t status;
L
Linus Torvalds 已提交
70
	atomic_t io_count;
R
Richard Kennedy 已提交
71
	struct bio *bio;
72
	unsigned long start_time;
73
	spinlock_t endio_lock;
M
Mikulas Patocka 已提交
74
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
75 76
};

77 78
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
79 80 81
/*
 * Bits for the md->flags field.
 */
82
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
83
#define DMF_SUSPENDED 1
84
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
85
#define DMF_FREEING 3
86
#define DMF_DELETING 4
87
#define DMF_NOFLUSH_SUSPENDING 5
88 89
#define DMF_DEFERRED_REMOVE 6
#define DMF_SUSPENDED_INTERNALLY 7
L
Linus Torvalds 已提交
90

91 92
#define DM_NUMA_NODE NUMA_NO_NODE
static int dm_numa_node = DM_NUMA_NODE;
93

K
Kiyoshi Ueda 已提交
94 95 96 97 98 99 100 101
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
	struct bio_set *bs;
};

102 103 104 105 106 107
struct table_device {
	struct list_head list;
	atomic_t count;
	struct dm_dev dm_dev;
};

108
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
109
static struct kmem_cache *_rq_tio_cache;
110
static struct kmem_cache *_rq_cache;
111

112 113 114
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
115
#define RESERVED_BIO_BASED_IOS		16
116 117
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
static int __dm_get_module_param_int(int *module_param, int min, int max)
{
	int param = ACCESS_ONCE(*module_param);
	int modified_param = 0;
	bool modified = true;

	if (param < min)
		modified_param = min;
	else if (param > max)
		modified_param = max;
	else
		modified = false;

	if (modified) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
	}

	return param;
}

139 140
unsigned __dm_get_module_param(unsigned *module_param,
			       unsigned def, unsigned max)
141
{
142 143
	unsigned param = ACCESS_ONCE(*module_param);
	unsigned modified_param = 0;
144

145 146 147 148
	if (!param)
		modified_param = def;
	else if (param > max)
		modified_param = max;
149

150 151 152
	if (modified_param) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
153 154
	}

155
	return param;
156 157
}

158 159
unsigned dm_get_reserved_bio_based_ios(void)
{
160
	return __dm_get_module_param(&reserved_bio_based_ios,
161
				     RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS);
162 163 164
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

165 166 167 168 169 170
static unsigned dm_get_numa_node(void)
{
	return __dm_get_module_param_int(&dm_numa_node,
					 DM_NUMA_NODE, num_online_nodes() - 1);
}

L
Linus Torvalds 已提交
171 172
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
173
	int r = -ENOMEM;
L
Linus Torvalds 已提交
174 175

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
176
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
177
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
178
		return r;
L
Linus Torvalds 已提交
179

K
Kiyoshi Ueda 已提交
180 181
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
182
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
183

184
	_rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request),
185 186 187 188
				      __alignof__(struct request), 0, NULL);
	if (!_rq_cache)
		goto out_free_rq_tio_cache;

M
Mike Anderson 已提交
189
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
190
	if (r)
191
		goto out_free_rq_cache;
M
Mike Anderson 已提交
192

193 194 195 196 197 198
	deferred_remove_workqueue = alloc_workqueue("kdmremove", WQ_UNBOUND, 1);
	if (!deferred_remove_workqueue) {
		r = -ENOMEM;
		goto out_uevent_exit;
	}

L
Linus Torvalds 已提交
199 200
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
201
	if (r < 0)
202
		goto out_free_workqueue;
L
Linus Torvalds 已提交
203 204 205 206 207

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
208

209 210
out_free_workqueue:
	destroy_workqueue(deferred_remove_workqueue);
K
Kiyoshi Ueda 已提交
211 212
out_uevent_exit:
	dm_uevent_exit();
213 214
out_free_rq_cache:
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
215 216
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
217 218 219 220
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
221 222 223 224
}

static void local_exit(void)
{
M
Mikulas Patocka 已提交
225
	flush_scheduled_work();
226
	destroy_workqueue(deferred_remove_workqueue);
M
Mikulas Patocka 已提交
227

228
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
229
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
230
	kmem_cache_destroy(_io_cache);
231
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
232
	dm_uevent_exit();
L
Linus Torvalds 已提交
233 234 235 236 237 238

	_major = 0;

	DMINFO("cleaned up");
}

239
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
240 241 242 243
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
244
	dm_io_init,
245
	dm_kcopyd_init,
L
Linus Torvalds 已提交
246
	dm_interface_init,
M
Mikulas Patocka 已提交
247
	dm_statistics_init,
L
Linus Torvalds 已提交
248 249
};

250
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
251 252 253 254
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
255
	dm_io_exit,
256
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
257
	dm_interface_exit,
M
Mikulas Patocka 已提交
258
	dm_statistics_exit,
L
Linus Torvalds 已提交
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
};

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]();
288 289 290 291 292

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
293 294 295 296 297
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
298 299 300 301 302
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
303
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
304 305 306
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
307 308
	spin_lock(&_minor_lock);

A
Al Viro 已提交
309
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
310 311 312
	if (!md)
		goto out;

313
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
314
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
315 316 317 318
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
319
	dm_get(md);
320
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
321 322 323 324
out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
325 326
}

327
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
328
{
329
	struct mapped_device *md;
330

331 332
	spin_lock(&_minor_lock);

333 334 335 336
	md = disk->private_data;
	if (WARN_ON(!md))
		goto out;

M
Mikulas Patocka 已提交
337 338
	if (atomic_dec_and_test(&md->open_count) &&
	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
339
		queue_work(deferred_remove_workqueue, &deferred_remove_work);
M
Mikulas Patocka 已提交
340

L
Linus Torvalds 已提交
341
	dm_put(md);
342
out:
343
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
344 345
}

346 347 348 349 350 351 352 353
int dm_open_count(struct mapped_device *md)
{
	return atomic_read(&md->open_count);
}

/*
 * Guarantees nothing is using the device before it's deleted.
 */
M
Mikulas Patocka 已提交
354
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
355 356 357 358 359
{
	int r = 0;

	spin_lock(&_minor_lock);

M
Mikulas Patocka 已提交
360
	if (dm_open_count(md)) {
361
		r = -EBUSY;
M
Mikulas Patocka 已提交
362 363 364 365
		if (mark_deferred)
			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
		r = -EEXIST;
366 367 368 369 370 371 372 373
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

M
Mikulas Patocka 已提交
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
int dm_cancel_deferred_remove(struct mapped_device *md)
{
	int r = 0;

	spin_lock(&_minor_lock);

	if (test_bit(DMF_DELETING, &md->flags))
		r = -EBUSY;
	else
		clear_bit(DMF_DEFERRED_REMOVE, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

static void do_deferred_remove(struct work_struct *w)
{
	dm_deferred_remove();
}

M
Mikulas Patocka 已提交
395 396 397 398 399
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

400 401 402 403 404
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
	return md->queue;
}

M
Mikulas Patocka 已提交
405 406 407 408 409
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
410 411 412 413 414 415 416
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);
}

417 418 419
static int dm_grab_bdev_for_ioctl(struct mapped_device *md,
				  struct block_device **bdev,
				  fmode_t *mode)
420
{
421
	struct dm_target *tgt;
422
	struct dm_table *map;
423
	int srcu_idx, r;
424

425
retry:
C
Christoph Hellwig 已提交
426
	r = -ENOTTY;
427
	map = dm_get_live_table(md, &srcu_idx);
428 429 430 431 432 433 434
	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;

435 436
	tgt = dm_table_get_target(map, 0);
	if (!tgt->type->prepare_ioctl)
437
		goto out;
438

439
	if (dm_suspended_md(md)) {
440 441 442 443
		r = -EAGAIN;
		goto out;
	}

444
	r = tgt->type->prepare_ioctl(tgt, bdev, mode);
C
Christoph Hellwig 已提交
445 446
	if (r < 0)
		goto out;
447

448 449
	bdgrab(*bdev);
	dm_put_live_table(md, srcu_idx);
C
Christoph Hellwig 已提交
450
	return r;
451 452

out:
453
	dm_put_live_table(md, srcu_idx);
454
	if (r == -ENOTCONN && !fatal_signal_pending(current)) {
455 456 457
		msleep(10);
		goto retry;
	}
C
Christoph Hellwig 已提交
458 459 460 461 462 463 464
	return r;
}

static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
			unsigned int cmd, unsigned long arg)
{
	struct mapped_device *md = bdev->bd_disk->private_data;
465
	int r;
C
Christoph Hellwig 已提交
466

467
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
C
Christoph Hellwig 已提交
468 469
	if (r < 0)
		return r;
470

C
Christoph Hellwig 已提交
471 472
	if (r > 0) {
		/*
473 474
		 * Target determined this ioctl is being issued against a
		 * subset of the parent bdev; require extra privileges.
C
Christoph Hellwig 已提交
475
		 */
476 477 478 479 480
		if (!capable(CAP_SYS_RAWIO)) {
			DMWARN_LIMIT(
	"%s: sending ioctl %x to DM device without required privilege.",
				current->comm, cmd);
			r = -ENOIOCTLCMD;
C
Christoph Hellwig 已提交
481
			goto out;
482
		}
C
Christoph Hellwig 已提交
483
	}
484

485
	r =  __blkdev_driver_ioctl(bdev, mode, cmd, arg);
C
Christoph Hellwig 已提交
486
out:
487
	bdput(bdev);
488 489 490
	return r;
}

A
Alasdair G Kergon 已提交
491
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
492 493 494 495
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
496
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
497 498 499 500
{
	mempool_free(io, md->io_pool);
}

501
static void free_tio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
502
{
503
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
504 505
}

506
int md_in_flight(struct mapped_device *md)
K
Kiyoshi Ueda 已提交
507 508 509 510 511
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

512 513 514
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
515
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
516
	int cpu;
M
Mikulas Patocka 已提交
517
	int rw = bio_data_dir(bio);
518 519 520

	io->start_time = jiffies;

T
Tejun Heo 已提交
521 522 523
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
524 525
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
526 527

	if (unlikely(dm_stats_used(&md->stats)))
528 529 530
		dm_stats_account_io(&md->stats, bio_data_dir(bio),
				    bio->bi_iter.bi_sector, bio_sectors(bio),
				    false, 0, &io->stats_aux);
531 532
}

533
static void end_io_acct(struct dm_io *io)
534 535 536 537
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
538
	int pending;
539 540
	int rw = bio_data_dir(bio);

541
	generic_end_io_acct(rw, &dm_disk(md)->part0, io->start_time);
542

M
Mikulas Patocka 已提交
543
	if (unlikely(dm_stats_used(&md->stats)))
544 545 546
		dm_stats_account_io(&md->stats, bio_data_dir(bio),
				    bio->bi_iter.bi_sector, bio_sectors(bio),
				    true, duration, &io->stats_aux);
M
Mikulas Patocka 已提交
547

548 549
	/*
	 * After this is decremented the bio must not be touched if it is
550
	 * a flush.
551
	 */
552 553
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
554
	pending += atomic_read(&md->pending[rw^0x1]);
555

556 557 558
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
559 560
}

L
Linus Torvalds 已提交
561 562 563
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
564
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
565
{
566
	unsigned long flags;
L
Linus Torvalds 已提交
567

568
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
569
	bio_list_add(&md->deferred, bio);
570
	spin_unlock_irqrestore(&md->deferred_lock, flags);
571
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
572 573 574 575 576
}

/*
 * 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
577
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
578
 */
M
Mikulas Patocka 已提交
579
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
580
{
M
Mikulas Patocka 已提交
581 582 583 584
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
585

M
Mikulas Patocka 已提交
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
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
606

M
Mikulas Patocka 已提交
607 608 609
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
610 611
}

612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
/*
 * Open a table device so we can use it as a map destination.
 */
static int open_table_device(struct table_device *td, dev_t dev,
			     struct mapped_device *md)
{
	static char *_claim_ptr = "I belong to device-mapper";
	struct block_device *bdev;

	int r;

	BUG_ON(td->dm_dev.bdev);

	bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _claim_ptr);
	if (IS_ERR(bdev))
		return PTR_ERR(bdev);

	r = bd_link_disk_holder(bdev, dm_disk(md));
	if (r) {
		blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL);
		return r;
	}

	td->dm_dev.bdev = bdev;
636
	td->dm_dev.dax_dev = dax_get_by_host(bdev->bd_disk->disk_name);
637 638 639 640 641 642 643 644 645 646 647 648 649
	return 0;
}

/*
 * Close a table device that we've been using.
 */
static void close_table_device(struct table_device *td, struct mapped_device *md)
{
	if (!td->dm_dev.bdev)
		return;

	bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md));
	blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL);
650
	put_dax(td->dm_dev.dax_dev);
651
	td->dm_dev.bdev = NULL;
652
	td->dm_dev.dax_dev = NULL;
653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
}

static struct table_device *find_table_device(struct list_head *l, dev_t dev,
					      fmode_t mode) {
	struct table_device *td;

	list_for_each_entry(td, l, list)
		if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode)
			return td;

	return NULL;
}

int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode,
			struct dm_dev **result) {
	int r;
	struct table_device *td;

	mutex_lock(&md->table_devices_lock);
	td = find_table_device(&md->table_devices, dev, mode);
	if (!td) {
674
		td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id);
675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728
		if (!td) {
			mutex_unlock(&md->table_devices_lock);
			return -ENOMEM;
		}

		td->dm_dev.mode = mode;
		td->dm_dev.bdev = NULL;

		if ((r = open_table_device(td, dev, md))) {
			mutex_unlock(&md->table_devices_lock);
			kfree(td);
			return r;
		}

		format_dev_t(td->dm_dev.name, dev);

		atomic_set(&td->count, 0);
		list_add(&td->list, &md->table_devices);
	}
	atomic_inc(&td->count);
	mutex_unlock(&md->table_devices_lock);

	*result = &td->dm_dev;
	return 0;
}
EXPORT_SYMBOL_GPL(dm_get_table_device);

void dm_put_table_device(struct mapped_device *md, struct dm_dev *d)
{
	struct table_device *td = container_of(d, struct table_device, dm_dev);

	mutex_lock(&md->table_devices_lock);
	if (atomic_dec_and_test(&td->count)) {
		close_table_device(td, md);
		list_del(&td->list);
		kfree(td);
	}
	mutex_unlock(&md->table_devices_lock);
}
EXPORT_SYMBOL(dm_put_table_device);

static void free_table_devices(struct list_head *devices)
{
	struct list_head *tmp, *next;

	list_for_each_safe(tmp, next, devices) {
		struct table_device *td = list_entry(tmp, struct table_device, list);

		DMWARN("dm_destroy: %s still exists with %d references",
		       td->dm_dev.name, atomic_read(&td->count));
		kfree(td);
	}
}

D
Darrick J. Wong 已提交
729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
/*
 * 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
756 757 758 759 760 761 762 763 764
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

765 766 767 768 769
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
770 771 772 773
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
774
static void dec_pending(struct dm_io *io, blk_status_t error)
L
Linus Torvalds 已提交
775
{
776
	unsigned long flags;
777
	blk_status_t io_error;
778 779
	struct bio *bio;
	struct mapped_device *md = io->md;
780 781

	/* Push-back supersedes any I/O errors */
782 783
	if (unlikely(error)) {
		spin_lock_irqsave(&io->endio_lock, flags);
784 785 786
		if (!(io->status == BLK_STS_DM_REQUEUE &&
				__noflush_suspending(md)))
			io->status = error;
787 788
		spin_unlock_irqrestore(&io->endio_lock, flags);
	}
L
Linus Torvalds 已提交
789 790

	if (atomic_dec_and_test(&io->io_count)) {
791
		if (io->status == BLK_STS_DM_REQUEUE) {
792 793 794
			/*
			 * Target requested pushing back the I/O.
			 */
795
			spin_lock_irqsave(&md->deferred_lock, flags);
796 797 798
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
799
				/* noflush suspend was interrupted. */
800
				io->status = BLK_STS_IOERR;
801
			spin_unlock_irqrestore(&md->deferred_lock, flags);
802 803
		}

804
		io_error = io->status;
805
		bio = io->bio;
806 807 808
		end_io_acct(io);
		free_io(md, io);

809
		if (io_error == BLK_STS_DM_REQUEUE)
810
			return;
811

J
Jens Axboe 已提交
812
		if ((bio->bi_opf & REQ_PREFLUSH) && bio->bi_iter.bi_size) {
813
			/*
814
			 * Preflush done for flush with data, reissue
815
			 * without REQ_PREFLUSH.
816
			 */
J
Jens Axboe 已提交
817
			bio->bi_opf &= ~REQ_PREFLUSH;
818
			queue_io(md, bio);
819
		} else {
820
			/* done with normal IO or empty flush */
821
			bio->bi_status = io_error;
822
			bio_endio(bio);
823
		}
L
Linus Torvalds 已提交
824 825 826
	}
}

827
void disable_write_same(struct mapped_device *md)
828 829 830 831 832 833 834
{
	struct queue_limits *limits = dm_get_queue_limits(md);

	/* device doesn't really support WRITE SAME, disable it */
	limits->max_write_same_sectors = 0;
}

835 836 837 838 839 840 841 842
void disable_write_zeroes(struct mapped_device *md)
{
	struct queue_limits *limits = dm_get_queue_limits(md);

	/* device doesn't really support WRITE ZEROES, disable it */
	limits->max_write_zeroes_sectors = 0;
}

843
static void clone_endio(struct bio *bio)
L
Linus Torvalds 已提交
844
{
845
	blk_status_t error = bio->bi_status;
M
Mikulas Patocka 已提交
846
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
847
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
848
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
849 850
	dm_endio_fn endio = tio->ti->type->end_io;

851
	if (unlikely(error == BLK_STS_TARGET)) {
852 853 854 855 856 857 858
		if (bio_op(bio) == REQ_OP_WRITE_SAME &&
		    !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors)
			disable_write_same(md);
		if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
		    !bdev_get_queue(bio->bi_bdev)->limits.max_write_zeroes_sectors)
			disable_write_zeroes(md);
	}
859

860
	if (endio) {
861
		int r = endio(tio->ti, bio, &error);
862 863
		switch (r) {
		case DM_ENDIO_REQUEUE:
864
			error = BLK_STS_DM_REQUEUE;
865 866 867 868 869 870 871 872 873 874 875 876
			/*FALLTHRU*/
		case DM_ENDIO_DONE:
			break;
		case DM_ENDIO_INCOMPLETE:
			/* The target will handle the io */
			return;
		default:
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
	}

877
	free_tio(tio);
878
	dec_pending(io, error);
L
Linus Torvalds 已提交
879 880
}

881 882 883 884 885 886 887 888 889 890 891 892
/*
 * 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
893
{
894
	sector_t len = max_io_len_target_boundary(sector, ti);
895
	sector_t offset, max_len;
L
Linus Torvalds 已提交
896 897

	/*
898
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
899
	 */
900 901 902 903 904 905 906 907 908 909
	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
910 911 912 913 914
	}

	return len;
}

915 916 917 918 919 920 921 922 923 924 925 926 927 928 929
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);

930 931
static struct dm_target *dm_dax_get_live_target(struct mapped_device *md,
		sector_t sector, int *srcu_idx)
932 933 934 935
{
	struct dm_table *map;
	struct dm_target *ti;

936
	map = dm_get_live_table(md, srcu_idx);
937
	if (!map)
938
		return NULL;
939 940 941

	ti = dm_table_find_target(map, sector);
	if (!dm_target_is_valid(ti))
942
		return NULL;
943

944 945
	return ti;
}
946

947 948 949 950 951 952 953 954
static long dm_dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff,
		long nr_pages, void **kaddr, pfn_t *pfn)
{
	struct mapped_device *md = dax_get_private(dax_dev);
	sector_t sector = pgoff * PAGE_SECTORS;
	struct dm_target *ti;
	long len, ret = -EIO;
	int srcu_idx;
955

956
	ti = dm_dax_get_live_target(md, sector, &srcu_idx);
957

958 959 960 961 962 963 964 965
	if (!ti)
		goto out;
	if (!ti->type->direct_access)
		goto out;
	len = max_io_len(sector, ti) / PAGE_SECTORS;
	if (len < 1)
		goto out;
	nr_pages = min(len, nr_pages);
966
	if (ti->type->direct_access)
967 968
		ret = ti->type->direct_access(ti, pgoff, nr_pages, kaddr, pfn);

969
 out:
970
	dm_put_live_table(md, srcu_idx);
971 972

	return ret;
973 974
}

975 976
/*
 * A target may call dm_accept_partial_bio only from the map routine.  It is
977
 * allowed for all bio types except REQ_PREFLUSH.
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
 *
 * dm_accept_partial_bio informs the dm that the target only wants to process
 * additional n_sectors sectors of the bio and the rest of the data should be
 * sent in a next bio.
 *
 * A diagram that explains the arithmetics:
 * +--------------------+---------------+-------+
 * |         1          |       2       |   3   |
 * +--------------------+---------------+-------+
 *
 * <-------------- *tio->len_ptr --------------->
 *                      <------- bi_size ------->
 *                      <-- n_sectors -->
 *
 * Region 1 was already iterated over with bio_advance or similar function.
 *	(it may be empty if the target doesn't use bio_advance)
 * Region 2 is the remaining bio size that the target wants to process.
 *	(it may be empty if region 1 is non-empty, although there is no reason
 *	 to make it empty)
 * The target requires that region 3 is to be sent in the next bio.
 *
 * If the target wants to receive multiple copies of the bio (via num_*bios, etc),
 * the partially processed part (the sum of regions 1+2) must be the same for all
 * copies of the bio.
 */
void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors)
{
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
	unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT;
J
Jens Axboe 已提交
1007
	BUG_ON(bio->bi_opf & REQ_PREFLUSH);
1008 1009 1010 1011 1012 1013 1014
	BUG_ON(bi_size > *tio->len_ptr);
	BUG_ON(n_sectors > bi_size);
	*tio->len_ptr -= bi_size - n_sectors;
	bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT;
}
EXPORT_SYMBOL_GPL(dm_accept_partial_bio);

1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
/*
 * Flush current->bio_list when the target map method blocks.
 * This fixes deadlocks in snapshot and possibly in other targets.
 */
struct dm_offload {
	struct blk_plug plug;
	struct blk_plug_cb cb;
};

static void flush_current_bio_list(struct blk_plug_cb *cb, bool from_schedule)
{
	struct dm_offload *o = container_of(cb, struct dm_offload, cb);
	struct bio_list list;
	struct bio *bio;
1029
	int i;
1030 1031 1032 1033 1034 1035

	INIT_LIST_HEAD(&o->cb.list);

	if (unlikely(!current->bio_list))
		return;

1036 1037 1038 1039 1040 1041
	for (i = 0; i < 2; i++) {
		list = current->bio_list[i];
		bio_list_init(&current->bio_list[i]);

		while ((bio = bio_list_pop(&list))) {
			struct bio_set *bs = bio->bi_pool;
1042 1043
			if (unlikely(!bs) || bs == fs_bio_set ||
			    !bs->rescue_workqueue) {
1044 1045 1046 1047 1048 1049 1050 1051
				bio_list_add(&current->bio_list[i], bio);
				continue;
			}

			spin_lock(&bs->rescue_lock);
			bio_list_add(&bs->rescue_list, bio);
			queue_work(bs->rescue_workqueue, &bs->rescue_work);
			spin_unlock(&bs->rescue_lock);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
		}
	}
}

static void dm_offload_start(struct dm_offload *o)
{
	blk_start_plug(&o->plug);
	o->cb.callback = flush_current_bio_list;
	list_add(&o->cb.list, &current->plug->cb_list);
}

static void dm_offload_end(struct dm_offload *o)
{
	list_del(&o->cb.list);
	blk_finish_plug(&o->plug);
}

A
Alasdair G Kergon 已提交
1069
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1070 1071
{
	int r;
1072
	sector_t sector;
1073
	struct dm_offload o;
1074
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1075
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1076 1077 1078 1079 1080 1081 1082 1083 1084

	clone->bi_end_io = clone_endio;

	/*
	 * 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);
1085
	sector = clone->bi_iter.bi_sector;
1086 1087

	dm_offload_start(&o);
M
Mikulas Patocka 已提交
1088
	r = ti->type->map(ti, clone);
1089 1090
	dm_offload_end(&o);

1091 1092 1093 1094
	switch (r) {
	case DM_MAPIO_SUBMITTED:
		break;
	case DM_MAPIO_REMAPPED:
L
Linus Torvalds 已提交
1095
		/* the bio has been remapped so dispatch it */
1096 1097
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
L
Linus Torvalds 已提交
1098
		generic_make_request(clone);
1099 1100
		break;
	case DM_MAPIO_KILL:
1101 1102 1103
		dec_pending(tio->io, BLK_STS_IOERR);
		free_tio(tio);
		break;
1104
	case DM_MAPIO_REQUEUE:
1105
		dec_pending(tio->io, BLK_STS_DM_REQUEUE);
1106
		free_tio(tio);
1107 1108
		break;
	default:
1109 1110
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1111 1112 1113 1114 1115 1116 1117 1118 1119
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1120
	unsigned sector_count;
L
Linus Torvalds 已提交
1121 1122
};

1123
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1124
{
1125 1126
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1127 1128 1129 1130 1131
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1132 1133
static int clone_bio(struct dm_target_io *tio, struct bio *bio,
		     sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1134
{
1135
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1136

1137 1138
	__bio_clone_fast(clone, bio);

1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
	if (unlikely(bio_integrity(bio) != NULL)) {
		int r;

		if (unlikely(!dm_target_has_integrity(tio->ti->type) &&
			     !dm_target_passes_integrity(tio->ti->type))) {
			DMWARN("%s: the target %s doesn't support integrity data.",
				dm_device_name(tio->io->md),
				tio->ti->type->name);
			return -EIO;
		}

		r = bio_integrity_clone(clone, bio, GFP_NOIO);
1151 1152 1153
		if (r < 0)
			return r;
	}
A
Alasdair G Kergon 已提交
1154

1155 1156 1157
	bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
	clone->bi_iter.bi_size = to_bytes(len);

1158
	if (unlikely(bio_integrity(bio) != NULL))
1159
		bio_integrity_trim(clone, 0, len);
1160 1161

	return 0;
L
Linus Torvalds 已提交
1162 1163
}

1164
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1165
				      struct dm_target *ti,
1166
				      unsigned target_bio_nr)
1167
{
1168 1169 1170
	struct dm_target_io *tio;
	struct bio *clone;

1171
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1172
	tio = container_of(clone, struct dm_target_io, clone);
1173 1174 1175

	tio->io = ci->io;
	tio->ti = ti;
1176
	tio->target_bio_nr = target_bio_nr;
1177 1178 1179 1180

	return tio;
}

1181 1182
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1183
				       unsigned target_bio_nr, unsigned *len)
1184
{
1185
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1186
	struct bio *clone = &tio->clone;
1187

1188 1189
	tio->len_ptr = len;

1190
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1191
	if (len)
1192
		bio_setup_sector(clone, ci->sector, *len);
1193

A
Alasdair G Kergon 已提交
1194
	__map_bio(tio);
1195 1196
}

1197
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1198
				  unsigned num_bios, unsigned *len)
1199
{
1200
	unsigned target_bio_nr;
1201

1202
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1203
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1204 1205
}

1206
static int __send_empty_flush(struct clone_info *ci)
1207
{
1208
	unsigned target_nr = 0;
1209 1210
	struct dm_target *ti;

1211
	BUG_ON(bio_has_data(ci->bio));
1212
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1213
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1214 1215 1216 1217

	return 0;
}

1218
static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1219
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1220
{
1221
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1222
	struct dm_target_io *tio;
1223 1224
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
1225
	int r = 0;
M
Mike Snitzer 已提交
1226

1227 1228 1229 1230 1231
	/*
	 * 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
1232

1233
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1234
		tio = alloc_tio(ci, ti, target_bio_nr);
1235
		tio->len_ptr = len;
1236
		r = clone_bio(tio, bio, sector, *len);
1237
		if (r < 0) {
1238
			free_tio(tio);
1239
			break;
1240
		}
1241 1242
		__map_bio(tio);
	}
1243 1244

	return r;
M
Mike Snitzer 已提交
1245 1246
}

1247
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1248

1249
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1250
{
1251
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1252 1253
}

1254
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1255
{
1256
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1257 1258
}

1259 1260 1261 1262 1263
static unsigned get_num_write_zeroes_bios(struct dm_target *ti)
{
	return ti->num_write_zeroes_bios;
}

M
Mike Snitzer 已提交
1264
typedef bool (*is_split_required_fn)(struct dm_target *ti);
1265

M
Mike Snitzer 已提交
1266 1267
static bool is_split_required_for_discard(struct dm_target *ti)
{
1268
	return ti->split_discard_bios;
1269 1270
}

1271 1272 1273
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
1274
{
M
Mike Snitzer 已提交
1275
	struct dm_target *ti;
1276
	unsigned len;
1277
	unsigned num_bios;
1278

1279 1280 1281 1282
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
1283

M
Mike Snitzer 已提交
1284
		/*
M
Mike Snitzer 已提交
1285 1286
		 * 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
1287
		 * reconfiguration might also have changed that since the
1288
		 * check was performed.
M
Mike Snitzer 已提交
1289
		 */
1290 1291
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1292
			return -EOPNOTSUPP;
1293

M
Mike Snitzer 已提交
1294
		if (is_split_required && !is_split_required(ti))
1295
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1296
		else
1297
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1298

1299
		__send_duplicate_bios(ci, ti, num_bios, &len);
1300

1301 1302
		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1303 1304

	return 0;
1305 1306
}

1307
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1308
{
1309 1310
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1311
}
1312

1313
static int __send_write_same(struct clone_info *ci)
1314
{
1315
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
1316 1317
}

1318 1319 1320 1321 1322
static int __send_write_zeroes(struct clone_info *ci)
{
	return __send_changing_extent_only(ci, get_num_write_zeroes_bios, NULL);
}

A
Alasdair G Kergon 已提交
1323 1324 1325
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1326
static int __split_and_process_non_flush(struct clone_info *ci)
1327
{
1328
	struct bio *bio = ci->bio;
1329
	struct dm_target *ti;
1330
	unsigned len;
1331
	int r;
1332

M
Mike Christie 已提交
1333
	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
1334
		return __send_discard(ci);
M
Mike Christie 已提交
1335
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME))
1336
		return __send_write_same(ci);
1337 1338
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES))
		return __send_write_zeroes(ci);
1339

1340 1341 1342 1343
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1344
	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
1345

1346 1347 1348
	r = __clone_and_map_data_bio(ci, ti, ci->sector, &len);
	if (r < 0)
		return r;
1349

1350 1351
	ci->sector += len;
	ci->sector_count -= len;
1352

1353
	return 0;
1354 1355
}

L
Linus Torvalds 已提交
1356
/*
1357
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1358
 */
M
Mikulas Patocka 已提交
1359 1360
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
1361
{
L
Linus Torvalds 已提交
1362
	struct clone_info ci;
1363
	int error = 0;
L
Linus Torvalds 已提交
1364

M
Mikulas Patocka 已提交
1365
	if (unlikely(!map)) {
1366
		bio_io_error(bio);
1367 1368
		return;
	}
1369

M
Mikulas Patocka 已提交
1370
	ci.map = map;
L
Linus Torvalds 已提交
1371 1372
	ci.md = md;
	ci.io = alloc_io(md);
1373
	ci.io->status = 0;
L
Linus Torvalds 已提交
1374 1375 1376
	atomic_set(&ci.io->io_count, 1);
	ci.io->bio = bio;
	ci.io->md = md;
1377
	spin_lock_init(&ci.io->endio_lock);
1378
	ci.sector = bio->bi_iter.bi_sector;
1379

1380
	start_io_acct(ci.io);
1381

J
Jens Axboe 已提交
1382
	if (bio->bi_opf & REQ_PREFLUSH) {
1383 1384
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1385
		error = __send_empty_flush(&ci);
1386 1387
		/* dec_pending submits any data associated with flush */
	} else {
1388
		ci.bio = bio;
1389
		ci.sector_count = bio_sectors(bio);
1390
		while (ci.sector_count && !error)
1391
			error = __split_and_process_non_flush(&ci);
1392
	}
1393

L
Linus Torvalds 已提交
1394
	/* drop the extra reference count */
1395
	dec_pending(ci.io, error);
1396
}
L
Linus Torvalds 已提交
1397 1398 1399
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/
1400

1401
/*
L
Linus Torvalds 已提交
1402 1403
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
1404
 */
1405
static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio)
1406
{
1407
	int rw = bio_data_dir(bio);
1408
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1409 1410
	int srcu_idx;
	struct dm_table *map;
1411

M
Mikulas Patocka 已提交
1412
	map = dm_get_live_table(md, &srcu_idx);
1413

1414
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1415

1416 1417
	/* 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
1418
		dm_put_live_table(md, srcu_idx);
1419

J
Jens Axboe 已提交
1420
		if (!(bio->bi_opf & REQ_RAHEAD))
1421 1422
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1423
			bio_io_error(bio);
1424
		return BLK_QC_T_NONE;
1425
	}
L
Linus Torvalds 已提交
1426

M
Mikulas Patocka 已提交
1427 1428
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1429
	return BLK_QC_T_NONE;
1430 1431
}

L
Linus Torvalds 已提交
1432 1433
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1434 1435 1436
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1437

1438
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mike Snitzer 已提交
1439
		if (dm_request_based(md)) {
1440
			/*
M
Mike Snitzer 已提交
1441 1442
			 * With request-based DM we only need to check the
			 * top-level queue for congestion.
1443
			 */
1444
			r = md->queue->backing_dev_info->wb.state & bdi_bits;
M
Mike Snitzer 已提交
1445 1446 1447
		} else {
			map = dm_get_live_table_fast(md);
			if (map)
1448
				r = dm_table_any_congested(map, bdi_bits);
M
Mike Snitzer 已提交
1449
			dm_put_live_table_fast(md);
1450 1451 1452
		}
	}

L
Linus Torvalds 已提交
1453 1454 1455 1456 1457 1458
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1459
static void free_minor(int minor)
L
Linus Torvalds 已提交
1460
{
1461
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1462
	idr_remove(&_minor_idr, minor);
1463
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1464 1465 1466 1467 1468
}

/*
 * See if the device with a specific minor # is free.
 */
1469
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1470
{
T
Tejun Heo 已提交
1471
	int r;
L
Linus Torvalds 已提交
1472 1473 1474 1475

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

T
Tejun Heo 已提交
1476
	idr_preload(GFP_KERNEL);
1477
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1478

T
Tejun Heo 已提交
1479
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1480

1481
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1482 1483 1484 1485
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1486 1487
}

1488
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1489
{
T
Tejun Heo 已提交
1490
	int r;
J
Jeff Mahoney 已提交
1491

T
Tejun Heo 已提交
1492
	idr_preload(GFP_KERNEL);
1493
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1494

T
Tejun Heo 已提交
1495
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1496

1497
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1498 1499 1500 1501 1502
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1503 1504
}

1505
static const struct block_device_operations dm_blk_dops;
1506
static const struct dax_operations dm_dax_ops;
L
Linus Torvalds 已提交
1507

1508 1509
static void dm_wq_work(struct work_struct *work);

1510
void dm_init_md_queue(struct mapped_device *md)
1511 1512 1513
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
1514
	 * devices.  The type of this dm device may not have been decided yet.
1515 1516 1517 1518 1519 1520 1521
	 * 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);
1522 1523 1524 1525 1526 1527

	/*
	 * Initialize data that will only be used by a non-blk-mq DM queue
	 * - must do so here (in alloc_dev callchain) before queue is used
	 */
	md->queue->queuedata = md;
1528
	md->queue->backing_dev_info->congested_data = md;
1529
}
1530

1531
void dm_init_normal_md_queue(struct mapped_device *md)
1532
{
1533
	md->use_blk_mq = false;
1534 1535 1536 1537 1538
	dm_init_md_queue(md);

	/*
	 * Initialize aspects of queue that aren't relevant for blk-mq
	 */
1539
	md->queue->backing_dev_info->congested_fn = dm_any_congested;
1540 1541 1542
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

1543 1544 1545 1546 1547 1548
static void cleanup_mapped_device(struct mapped_device *md)
{
	if (md->wq)
		destroy_workqueue(md->wq);
	if (md->kworker_task)
		kthread_stop(md->kworker_task);
1549
	mempool_destroy(md->io_pool);
1550 1551 1552
	if (md->bs)
		bioset_free(md->bs);

1553 1554 1555 1556 1557 1558
	if (md->dax_dev) {
		kill_dax(md->dax_dev);
		put_dax(md->dax_dev);
		md->dax_dev = NULL;
	}

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569
	if (md->disk) {
		spin_lock(&_minor_lock);
		md->disk->private_data = NULL;
		spin_unlock(&_minor_lock);
		del_gendisk(md->disk);
		put_disk(md->disk);
	}

	if (md->queue)
		blk_cleanup_queue(md->queue);

1570 1571
	cleanup_srcu_struct(&md->io_barrier);

1572 1573 1574 1575
	if (md->bdev) {
		bdput(md->bdev);
		md->bdev = NULL;
	}
1576 1577

	dm_mq_cleanup_mapped_device(md);
1578 1579
}

L
Linus Torvalds 已提交
1580 1581 1582
/*
 * Allocate and initialise a blank device with a given minor.
 */
1583
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1584
{
1585
	int r, numa_node_id = dm_get_numa_node();
1586
	struct dax_device *dax_dev;
1587
	struct mapped_device *md;
1588
	void *old_md;
L
Linus Torvalds 已提交
1589

1590
	md = kzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595
	if (!md) {
		DMWARN("unable to allocate device, out of memory.");
		return NULL;
	}

1596
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1597
		goto bad_module_get;
1598

L
Linus Torvalds 已提交
1599
	/* get a minor number for the dev */
1600
	if (minor == DM_ANY_MINOR)
1601
		r = next_free_minor(&minor);
1602
	else
1603
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1604
	if (r < 0)
M
Milan Broz 已提交
1605
		goto bad_minor;
L
Linus Torvalds 已提交
1606

M
Mikulas Patocka 已提交
1607 1608 1609 1610
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

1611
	md->numa_node_id = numa_node_id;
1612
	md->use_blk_mq = dm_use_blk_mq_default();
1613
	md->init_tio_pdu = false;
1614
	md->type = DM_TYPE_NONE;
1615
	mutex_init(&md->suspend_lock);
1616
	mutex_init(&md->type_lock);
1617
	mutex_init(&md->table_devices_lock);
1618
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1619
	atomic_set(&md->holders, 1);
1620
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1621
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1622 1623
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
1624
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
1625
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1626

1627
	md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1628
	if (!md->queue)
1629
		goto bad;
L
Linus Torvalds 已提交
1630

1631
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1632

1633
	md->disk = alloc_disk_node(1, numa_node_id);
L
Linus Torvalds 已提交
1634
	if (!md->disk)
1635
		goto bad;
L
Linus Torvalds 已提交
1636

1637 1638
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1639
	init_waitqueue_head(&md->wait);
1640
	INIT_WORK(&md->work, dm_wq_work);
1641
	init_waitqueue_head(&md->eventq);
1642
	init_completion(&md->kobj_holder.completion);
1643
	md->kworker_task = NULL;
1644

L
Linus Torvalds 已提交
1645 1646 1647 1648 1649 1650
	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);
1651 1652 1653 1654 1655 1656

	dax_dev = alloc_dax(md, md->disk->disk_name, &dm_dax_ops);
	if (!dax_dev)
		goto bad;
	md->dax_dev = dax_dev;

L
Linus Torvalds 已提交
1657
	add_disk(md->disk);
M
Mike Anderson 已提交
1658
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1659

T
Tejun Heo 已提交
1660
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
1661
	if (!md->wq)
1662
		goto bad;
1663

M
Mikulas Patocka 已提交
1664 1665
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
1666
		goto bad;
M
Mikulas Patocka 已提交
1667

1668
	bio_init(&md->flush_bio, NULL, 0);
1669
	md->flush_bio.bi_bdev = md->bdev;
J
Jan Kara 已提交
1670
	md->flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC;
1671

M
Mikulas Patocka 已提交
1672 1673
	dm_stats_init(&md->stats);

1674
	/* Populate the mapping, nobody knows we exist yet */
1675
	spin_lock(&_minor_lock);
1676
	old_md = idr_replace(&_minor_idr, md, minor);
1677
	spin_unlock(&_minor_lock);
1678 1679 1680

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1681 1682
	return md;

1683 1684
bad:
	cleanup_mapped_device(md);
M
Mikulas Patocka 已提交
1685
bad_io_barrier:
L
Linus Torvalds 已提交
1686
	free_minor(minor);
M
Milan Broz 已提交
1687
bad_minor:
1688
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1689
bad_module_get:
L
Linus Torvalds 已提交
1690 1691 1692 1693
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1694 1695
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1696 1697
static void free_dev(struct mapped_device *md)
{
1698
	int minor = MINOR(disk_devt(md->disk));
1699

M
Mikulas Patocka 已提交
1700
	unlock_fs(md);
1701

1702
	cleanup_mapped_device(md);
1703

1704
	free_table_devices(&md->table_devices);
1705 1706 1707
	dm_stats_cleanup(&md->stats);
	free_minor(minor);

1708
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1709 1710 1711
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1712 1713
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1714
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1715

1716 1717
	if (md->bs) {
		/* The md already has necessary mempools. */
1718
		if (dm_table_bio_based(t)) {
1719 1720 1721 1722 1723 1724 1725 1726
			/*
			 * 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;
		}
1727 1728 1729 1730 1731 1732 1733 1734 1735
		/*
		 * 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.
		 */
		goto out;
M
Mikulas Patocka 已提交
1736
	}
K
Kiyoshi Ueda 已提交
1737

1738
	BUG_ON(!p || md->io_pool || md->bs);
1739

K
Kiyoshi Ueda 已提交
1740 1741 1742 1743
	md->io_pool = p->io_pool;
	p->io_pool = NULL;
	md->bs = p->bs;
	p->bs = NULL;
1744

K
Kiyoshi Ueda 已提交
1745
out:
1746
	/* mempool bind completed, no longer need any mempools in the table */
K
Kiyoshi Ueda 已提交
1747 1748 1749
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
1750 1751 1752 1753 1754
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
1755 1756
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
1757 1758
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
1759 1760 1761 1762
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

1763
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
1764

L
Linus Torvalds 已提交
1765
	atomic_inc(&md->event_nr);
1766
	atomic_inc(&dm_global_event_nr);
L
Linus Torvalds 已提交
1767
	wake_up(&md->eventq);
1768
	wake_up(&dm_global_eventq);
L
Linus Torvalds 已提交
1769 1770
}

1771 1772 1773
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
1774
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
1775
{
1776 1777
	lockdep_assert_held(&md->suspend_lock);

1778
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
1779

1780
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
1781 1782
}

1783 1784 1785 1786 1787
/*
 * 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
1788
{
1789
	struct dm_table *old_map;
1790
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
1791 1792
	sector_t size;

1793 1794
	lockdep_assert_held(&md->suspend_lock);

L
Linus Torvalds 已提交
1795
	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
1796 1797 1798 1799

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
1800
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
1801 1802
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
1803
	__set_size(md, size);
1804

1805 1806
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
1807 1808 1809 1810 1811 1812 1813
	/*
	 * 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.
	 */
M
Mike Snitzer 已提交
1814
	if (dm_table_request_based(t)) {
1815
		dm_stop_queue(q);
M
Mike Snitzer 已提交
1816 1817 1818 1819 1820 1821 1822
		/*
		 * Leverage the fact that request-based DM targets are
		 * immutable singletons and establish md->immutable_target
		 * - used to optimize both dm_request_fn and dm_mq_queue_rq
		 */
		md->immutable_target = dm_table_get_immutable_target(t);
	}
K
Kiyoshi Ueda 已提交
1823 1824 1825

	__bind_mempools(md, t);

1826
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
1827
	rcu_assign_pointer(md->map, (void *)t);
1828 1829
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

1830
	dm_table_set_restrictions(t, q, limits);
1831 1832
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
1833

1834
	return old_map;
L
Linus Torvalds 已提交
1835 1836
}

1837 1838 1839 1840
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
1841
{
1842
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
1843 1844

	if (!map)
1845
		return NULL;
L
Linus Torvalds 已提交
1846 1847

	dm_table_event_callback(map, NULL, NULL);
1848
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
1849
	dm_sync_table(md);
1850 1851

	return map;
L
Linus Torvalds 已提交
1852 1853 1854 1855 1856
}

/*
 * Constructor for a new device.
 */
1857
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
1858 1859 1860
{
	struct mapped_device *md;

1861
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
1862 1863 1864
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
1865 1866
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
1867 1868 1869 1870
	*result = md;
	return 0;
}

1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
/*
 * 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);
}

1885
void dm_set_md_type(struct mapped_device *md, enum dm_queue_mode type)
1886
{
1887
	BUG_ON(!mutex_is_locked(&md->type_lock));
1888 1889 1890
	md->type = type;
}

1891
enum dm_queue_mode dm_get_md_type(struct mapped_device *md)
1892 1893 1894 1895
{
	return md->type;
}

1896 1897 1898 1899 1900
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
/*
 * 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);

1912 1913 1914
/*
 * Setup the DM device's queue based on md's type
 */
1915
int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t)
1916
{
1917
	int r;
1918
	enum dm_queue_mode type = dm_get_md_type(md);
1919

1920
	switch (type) {
1921
	case DM_TYPE_REQUEST_BASED:
1922
		r = dm_old_init_request_queue(md, t);
1923
		if (r) {
1924
			DMERR("Cannot initialize queue for request-based mapped device");
1925
			return r;
1926
		}
1927 1928
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
1929
		r = dm_mq_init_request_queue(md, t);
1930
		if (r) {
1931
			DMERR("Cannot initialize queue for request-based dm-mq mapped device");
1932 1933 1934 1935
			return r;
		}
		break;
	case DM_TYPE_BIO_BASED:
1936
	case DM_TYPE_DAX_BIO_BASED:
1937
		dm_init_normal_md_queue(md);
1938
		blk_queue_make_request(md->queue, dm_make_request);
1939 1940 1941 1942 1943 1944
		/*
		 * DM handles splitting bios as needed.  Free the bio_split bioset
		 * since it won't be used (saves 1 process per bio-based DM device).
		 */
		bioset_free(md->queue->bio_split);
		md->queue->bio_split = NULL;
1945 1946 1947

		if (type == DM_TYPE_DAX_BIO_BASED)
			queue_flag_set_unlocked(QUEUE_FLAG_DAX, md->queue);
1948
		break;
1949 1950 1951
	case DM_TYPE_NONE:
		WARN_ON_ONCE(true);
		break;
1952 1953 1954 1955 1956
	}

	return 0;
}

1957
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
1958 1959 1960 1961 1962 1963 1964
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

1965
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1966 1967

	md = idr_find(&_minor_idr, minor);
1968 1969 1970 1971 1972 1973 1974 1975 1976
	if (md) {
		if ((md == MINOR_ALLOCED ||
		     (MINOR(disk_devt(dm_disk(md))) != minor) ||
		     dm_deleting_md(md) ||
		     test_bit(DMF_FREEING, &md->flags))) {
			md = NULL;
			goto out;
		}
		dm_get(md);
J
Jeff Mahoney 已提交
1977
	}
L
Linus Torvalds 已提交
1978

J
Jeff Mahoney 已提交
1979
out:
1980
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1981

1982 1983
	return md;
}
A
Alasdair G Kergon 已提交
1984
EXPORT_SYMBOL_GPL(dm_get_md);
1985

A
Alasdair G Kergon 已提交
1986
void *dm_get_mdptr(struct mapped_device *md)
1987
{
A
Alasdair G Kergon 已提交
1988
	return md->interface_ptr;
L
Linus Torvalds 已提交
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
}

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);
1999
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2000 2001
}

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
int dm_hold(struct mapped_device *md)
{
	spin_lock(&_minor_lock);
	if (test_bit(DMF_FREEING, &md->flags)) {
		spin_unlock(&_minor_lock);
		return -EBUSY;
	}
	dm_get(md);
	spin_unlock(&_minor_lock);
	return 0;
}
EXPORT_SYMBOL_GPL(dm_hold);

2015 2016 2017 2018 2019 2020
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2021
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2022
{
2023
	struct request_queue *q = dm_get_md_queue(md);
M
Mike Anderson 已提交
2024
	struct dm_table *map;
M
Mikulas Patocka 已提交
2025
	int srcu_idx;
L
Linus Torvalds 已提交
2026

2027
	might_sleep();
J
Jeff Mahoney 已提交
2028

2029
	spin_lock(&_minor_lock);
2030 2031 2032
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);
2033

2034
	blk_set_queue_dying(q);
2035

2036
	if (dm_request_based(md) && md->kworker_task)
P
Petr Mladek 已提交
2037
		kthread_flush_worker(&md->kworker);
2038

2039 2040 2041 2042 2043
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
2044
	map = dm_get_live_table(md, &srcu_idx);
2045 2046 2047
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2048
	}
M
Mikulas Patocka 已提交
2049 2050
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);
2051
	mutex_unlock(&md->suspend_lock);
M
Mikulas Patocka 已提交
2052

2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
	/*
	 * 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
2084
}
E
Edward Goggin 已提交
2085
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2086

2087
static int dm_wait_for_completion(struct mapped_device *md, long task_state)
2088 2089
{
	int r = 0;
2090
	DEFINE_WAIT(wait);
2091 2092

	while (1) {
2093
		prepare_to_wait(&md->wait, &wait, task_state);
2094

2095
		if (!md_in_flight(md))
2096 2097
			break;

2098
		if (signal_pending_state(task_state, current)) {
2099 2100 2101 2102 2103 2104
			r = -EINTR;
			break;
		}

		io_schedule();
	}
2105
	finish_wait(&md->wait, &wait);
2106

2107 2108 2109
	return r;
}

L
Linus Torvalds 已提交
2110 2111 2112
/*
 * Process the deferred bios
 */
2113
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2114
{
2115 2116
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2117
	struct bio *c;
M
Mikulas Patocka 已提交
2118 2119
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2120

M
Mikulas Patocka 已提交
2121
	map = dm_get_live_table(md, &srcu_idx);
2122

2123
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2124 2125 2126 2127
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2128
		if (!c)
A
Alasdair G Kergon 已提交
2129
			break;
2130

K
Kiyoshi Ueda 已提交
2131 2132
		if (dm_request_based(md))
			generic_make_request(c);
2133
		else
M
Mikulas Patocka 已提交
2134
			__split_and_process_bio(md, map, c);
2135
	}
M
Milan Broz 已提交
2136

M
Mikulas Patocka 已提交
2137
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2138 2139
}

2140
static void dm_queue_flush(struct mapped_device *md)
2141
{
2142
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2143
	smp_mb__after_atomic();
2144
	queue_work(md->wq, &md->work);
2145 2146
}

L
Linus Torvalds 已提交
2147
/*
2148
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2149
 */
2150
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2151
{
2152
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2153
	struct queue_limits limits;
2154
	int r;
L
Linus Torvalds 已提交
2155

2156
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2157 2158

	/* device must be suspended */
2159
	if (!dm_suspended_md(md))
2160
		goto out;
L
Linus Torvalds 已提交
2161

2162 2163 2164 2165 2166 2167 2168
	/*
	 * 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
2169
		live_map = dm_get_live_table_fast(md);
2170 2171
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2172
		dm_put_live_table_fast(md);
2173 2174
	}

2175 2176 2177 2178 2179 2180
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2181
	}
2182

2183
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2184

2185
out:
2186
	mutex_unlock(&md->suspend_lock);
2187
	return map;
L
Linus Torvalds 已提交
2188 2189 2190 2191 2192 2193
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2194
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2195
{
2196
	int r;
L
Linus Torvalds 已提交
2197 2198

	WARN_ON(md->frozen_sb);
2199

2200
	md->frozen_sb = freeze_bdev(md->bdev);
2201
	if (IS_ERR(md->frozen_sb)) {
2202
		r = PTR_ERR(md->frozen_sb);
2203 2204
		md->frozen_sb = NULL;
		return r;
2205 2206
	}

2207 2208
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2209 2210 2211
	return 0;
}

2212
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2213
{
2214 2215 2216
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2217
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2218
	md->frozen_sb = NULL;
2219
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2220 2221 2222
}

/*
2223 2224 2225 2226
 * @suspend_flags: DM_SUSPEND_LOCKFS_FLAG and/or DM_SUSPEND_NOFLUSH_FLAG
 * @task_state: e.g. TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE
 * @dmf_suspended_flag: DMF_SUSPENDED or DMF_SUSPENDED_INTERNALLY
 *
2227 2228 2229
 * If __dm_suspend returns 0, the device is completely quiescent
 * now. There is no request-processing activity. All new requests
 * are being added to md->deferred list.
2230
 */
2231
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
2232
			unsigned suspend_flags, long task_state,
2233
			int dmf_suspended_flag)
L
Linus Torvalds 已提交
2234
{
2235 2236 2237
	bool do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG;
	bool noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG;
	int r;
L
Linus Torvalds 已提交
2238

2239 2240
	lockdep_assert_held(&md->suspend_lock);

2241 2242 2243 2244 2245 2246
	/*
	 * 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);
2247 2248
	else
		pr_debug("%s: suspending with flush\n", dm_device_name(md));
2249

2250 2251 2252 2253
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
2254 2255
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2256
	/*
K
Kiyoshi Ueda 已提交
2257 2258 2259 2260
	 * 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
2261 2262 2263
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
2264 2265
		if (r) {
			dm_table_presuspend_undo_targets(map);
2266
			return r;
2267
		}
2268
	}
L
Linus Torvalds 已提交
2269 2270

	/*
2271 2272 2273 2274 2275 2276 2277
	 * 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
2278 2279 2280
	 * __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
2281
	 */
2282
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2283 2284
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2285

2286
	/*
2287 2288
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2289
	 */
2290
	if (dm_request_based(md)) {
2291
		dm_stop_queue(md->queue);
2292
		if (md->kworker_task)
P
Petr Mladek 已提交
2293
			kthread_flush_worker(&md->kworker);
2294
	}
2295

2296 2297
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2298
	/*
2299 2300 2301
	 * 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
2302
	 */
2303
	r = dm_wait_for_completion(md, task_state);
2304 2305
	if (!r)
		set_bit(dmf_suspended_flag, &md->flags);
L
Linus Torvalds 已提交
2306

2307
	if (noflush)
2308
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2309 2310
	if (map)
		synchronize_srcu(&md->io_barrier);
2311

L
Linus Torvalds 已提交
2312
	/* were we interrupted ? */
2313
	if (r < 0) {
2314
		dm_queue_flush(md);
M
Milan Broz 已提交
2315

2316
		if (dm_request_based(md))
2317
			dm_start_queue(md->queue);
2318

2319
		unlock_fs(md);
2320
		dm_table_presuspend_undo_targets(map);
2321
		/* pushback list is already flushed, so skip flush */
2322
	}
L
Linus Torvalds 已提交
2323

2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364
	return r;
}

/*
 * 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.
 */
/*
 * Suspend mechanism in request-based dm.
 *
 * 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.
 *
 * To abort suspend, start the request_queue.
 */
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
{
	struct dm_table *map = NULL;
	int r = 0;

retry:
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

	if (dm_suspended_md(md)) {
		r = -EINVAL;
		goto out_unlock;
	}

	if (dm_suspended_internally_md(md)) {
		/* already internally suspended, wait for internal resume */
		mutex_unlock(&md->suspend_lock);
		r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE);
		if (r)
			return r;
		goto retry;
	}

2365
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2366

2367
	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED);
2368 2369
	if (r)
		goto out_unlock;
2370

2371 2372
	dm_table_postsuspend_targets(map);

2373
out_unlock:
2374
	mutex_unlock(&md->suspend_lock);
2375
	return r;
L
Linus Torvalds 已提交
2376 2377
}

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
static int __dm_resume(struct mapped_device *md, struct dm_table *map)
{
	if (map) {
		int r = dm_table_resume_targets(map);
		if (r)
			return r;
	}

	dm_queue_flush(md);

	/*
	 * 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))
2394
		dm_start_queue(md->queue);
2395 2396 2397 2398 2399 2400

	unlock_fs(md);

	return 0;
}

L
Linus Torvalds 已提交
2401 2402
int dm_resume(struct mapped_device *md)
{
2403
	int r;
2404
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2405

2406
retry:
2407
	r = -EINVAL;
2408 2409
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

2410
	if (!dm_suspended_md(md))
2411 2412
		goto out;

2413 2414 2415 2416 2417 2418 2419 2420 2421
	if (dm_suspended_internally_md(md)) {
		/* already internally suspended, wait for internal resume */
		mutex_unlock(&md->suspend_lock);
		r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE);
		if (r)
			return r;
		goto retry;
	}

2422
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2423
	if (!map || !dm_table_get_size(map))
2424
		goto out;
L
Linus Torvalds 已提交
2425

2426
	r = __dm_resume(md, map);
2427 2428
	if (r)
		goto out;
2429 2430

	clear_bit(DMF_SUSPENDED, &md->flags);
2431
out:
2432
	mutex_unlock(&md->suspend_lock);
2433

2434
	return r;
L
Linus Torvalds 已提交
2435 2436
}

M
Mikulas Patocka 已提交
2437 2438 2439 2440 2441 2442
/*
 * 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.
 */

2443
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
2444
{
2445 2446
	struct dm_table *map = NULL;

2447 2448
	lockdep_assert_held(&md->suspend_lock);

2449
	if (md->internal_suspend_count++)
2450 2451 2452 2453 2454 2455 2456
		return; /* nested internal suspend */

	if (dm_suspended_md(md)) {
		set_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
		return; /* nest suspend */
	}

2457
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2458 2459 2460 2461 2462 2463 2464

	/*
	 * Using TASK_UNINTERRUPTIBLE because only NOFLUSH internal suspend is
	 * supported.  Properly supporting a TASK_INTERRUPTIBLE internal suspend
	 * would require changing .presuspend to return an error -- avoid this
	 * until there is a need for more elaborate variants of internal suspend.
	 */
2465 2466
	(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE,
			    DMF_SUSPENDED_INTERNALLY);
2467 2468 2469 2470 2471 2472

	dm_table_postsuspend_targets(map);
}

static void __dm_internal_resume(struct mapped_device *md)
{
2473 2474 2475
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
2476 2477
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
2478
	if (dm_suspended_md(md))
2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
		goto done; /* resume from nested suspend */

	/*
	 * NOTE: existing callers don't need to call dm_table_resume_targets
	 * (which may fail -- so best to avoid it for now by passing NULL map)
	 */
	(void) __dm_resume(md, NULL);

done:
	clear_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
	smp_mb__after_atomic();
	wake_up_bit(&md->flags, DMF_SUSPENDED_INTERNALLY);
}

void dm_internal_suspend_noflush(struct mapped_device *md)
{
	mutex_lock(&md->suspend_lock);
	__dm_internal_suspend(md, DM_SUSPEND_NOFLUSH_FLAG);
	mutex_unlock(&md->suspend_lock);
}
EXPORT_SYMBOL_GPL(dm_internal_suspend_noflush);

void dm_internal_resume(struct mapped_device *md)
{
	mutex_lock(&md->suspend_lock);
	__dm_internal_resume(md);
	mutex_unlock(&md->suspend_lock);
}
EXPORT_SYMBOL_GPL(dm_internal_resume);

/*
 * Fast variants of internal suspend/resume hold md->suspend_lock,
 * which prevents interaction with userspace-driven suspend.
 */

void dm_internal_suspend_fast(struct mapped_device *md)
{
	mutex_lock(&md->suspend_lock);
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
2518 2519 2520 2521 2522 2523 2524
		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);
}
2525
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
2526

2527
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
2528
{
2529
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
2530 2531 2532 2533 2534 2535 2536
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
2537
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
2538

L
Linus Torvalds 已提交
2539 2540 2541
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2542
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2543
		       unsigned cookie)
2544
{
M
Milan Broz 已提交
2545 2546 2547 2548
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2549
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2550 2551 2552
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2553 2554
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2555
	}
2556 2557
}

M
Mike Anderson 已提交
2558 2559 2560 2561 2562
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
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
2574 2575 2576 2577 2578 2579 2580 2581 2582
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
2583 2584 2585 2586 2587 2588 2589 2590
/*
 * 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;
}
2591
EXPORT_SYMBOL_GPL(dm_disk);
L
Linus Torvalds 已提交
2592

M
Milan Broz 已提交
2593 2594
struct kobject *dm_kobject(struct mapped_device *md)
{
2595
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
2596 2597 2598 2599 2600 2601
}

struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
{
	struct mapped_device *md;

2602
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
2603

2604
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2605
	    dm_deleting_md(md))
2606 2607
		return NULL;

M
Milan Broz 已提交
2608 2609 2610 2611
	dm_get(md);
	return md;
}

2612
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2613 2614 2615 2616
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2617 2618 2619 2620 2621
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
2622 2623 2624 2625 2626
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

2627 2628
int dm_suspended(struct dm_target *ti)
{
2629
	return dm_suspended_md(dm_table_get_md(ti->table));
2630 2631 2632
}
EXPORT_SYMBOL_GPL(dm_suspended);

2633 2634
int dm_noflush_suspending(struct dm_target *ti)
{
2635
	return __noflush_suspending(dm_table_get_md(ti->table));
2636 2637 2638
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

2639
struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type,
2640
					    unsigned integrity, unsigned per_io_data_size)
K
Kiyoshi Ueda 已提交
2641
{
2642
	struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id);
2643
	unsigned int pool_size = 0;
J
Jun'ichi Nomura 已提交
2644
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2645 2646

	if (!pools)
2647
		return NULL;
K
Kiyoshi Ueda 已提交
2648

2649 2650
	switch (type) {
	case DM_TYPE_BIO_BASED:
2651
	case DM_TYPE_DAX_BIO_BASED:
2652
		pool_size = dm_get_reserved_bio_based_ios();
2653
		front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
2654 2655 2656 2657
	
		pools->io_pool = mempool_create_slab_pool(pool_size, _io_cache);
		if (!pools->io_pool)
			goto out;
2658 2659 2660
		break;
	case DM_TYPE_REQUEST_BASED:
	case DM_TYPE_MQ_REQUEST_BASED:
2661
		pool_size = dm_get_reserved_rq_based_ios();
2662
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
2663
		/* per_io_data_size is used for blk-mq pdu at queue allocation */
2664 2665 2666 2667 2668
		break;
	default:
		BUG();
	}

2669
	pools->bs = bioset_create(pool_size, front_pad, BIOSET_NEED_RESCUER);
K
Kiyoshi Ueda 已提交
2670
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2671
		goto out;
K
Kiyoshi Ueda 已提交
2672

2673
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2674
		goto out;
2675

K
Kiyoshi Ueda 已提交
2676
	return pools;
2677 2678 2679

out:
	dm_free_md_mempools(pools);
2680

2681
	return NULL;
K
Kiyoshi Ueda 已提交
2682 2683 2684 2685 2686 2687 2688
}

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

2689
	mempool_destroy(pools->io_pool);
2690

K
Kiyoshi Ueda 已提交
2691 2692 2693 2694 2695 2696
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

2697 2698 2699 2700 2701 2702 2703 2704 2705
struct dm_pr {
	u64	old_key;
	u64	new_key;
	u32	flags;
	bool	fail_early;
};

static int dm_call_pr(struct block_device *bdev, iterate_devices_callout_fn fn,
		      void *data)
2706 2707
{
	struct mapped_device *md = bdev->bd_disk->private_data;
2708 2709 2710
	struct dm_table *table;
	struct dm_target *ti;
	int ret = -ENOTTY, srcu_idx;
2711

2712 2713 2714
	table = dm_get_live_table(md, &srcu_idx);
	if (!table || !dm_table_get_size(table))
		goto out;
2715

2716 2717 2718 2719
	/* We only support devices that have a single target */
	if (dm_table_get_num_targets(table) != 1)
		goto out;
	ti = dm_table_get_target(table, 0);
2720

2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
	ret = -EINVAL;
	if (!ti->type->iterate_devices)
		goto out;

	ret = ti->type->iterate_devices(ti, fn, data);
out:
	dm_put_live_table(md, srcu_idx);
	return ret;
}

/*
 * For register / unregister we need to manually call out to every path.
 */
static int __dm_pr_register(struct dm_target *ti, struct dm_dev *dev,
			    sector_t start, sector_t len, void *data)
{
	struct dm_pr *pr = data;
	const struct pr_ops *ops = dev->bdev->bd_disk->fops->pr_ops;

	if (!ops || !ops->pr_register)
		return -EOPNOTSUPP;
	return ops->pr_register(dev->bdev, pr->old_key, pr->new_key, pr->flags);
}

static int dm_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
			  u32 flags)
{
	struct dm_pr pr = {
		.old_key	= old_key,
		.new_key	= new_key,
		.flags		= flags,
		.fail_early	= true,
	};
	int ret;

	ret = dm_call_pr(bdev, __dm_pr_register, &pr);
	if (ret && new_key) {
		/* unregister all paths if we failed to register any path */
		pr.old_key = new_key;
		pr.new_key = 0;
		pr.flags = 0;
		pr.fail_early = false;
		dm_call_pr(bdev, __dm_pr_register, &pr);
	}

	return ret;
2767 2768 2769
}

static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2770
			 u32 flags)
2771 2772 2773 2774
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2775
	int r;
2776

2777
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2778 2779 2780 2781 2782 2783 2784 2785 2786
	if (r < 0)
		return r;

	ops = bdev->bd_disk->fops->pr_ops;
	if (ops && ops->pr_reserve)
		r = ops->pr_reserve(bdev, key, type, flags);
	else
		r = -EOPNOTSUPP;

2787
	bdput(bdev);
2788 2789 2790 2791 2792 2793 2794 2795
	return r;
}

static int dm_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2796
	int r;
2797

2798
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2799 2800 2801 2802 2803 2804 2805 2806 2807
	if (r < 0)
		return r;

	ops = bdev->bd_disk->fops->pr_ops;
	if (ops && ops->pr_release)
		r = ops->pr_release(bdev, key, type);
	else
		r = -EOPNOTSUPP;

2808
	bdput(bdev);
2809 2810 2811 2812
	return r;
}

static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2813
			 enum pr_type type, bool abort)
2814 2815 2816 2817
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2818
	int r;
2819

2820
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2821 2822 2823 2824 2825 2826 2827 2828 2829
	if (r < 0)
		return r;

	ops = bdev->bd_disk->fops->pr_ops;
	if (ops && ops->pr_preempt)
		r = ops->pr_preempt(bdev, old_key, new_key, type, abort);
	else
		r = -EOPNOTSUPP;

2830
	bdput(bdev);
2831 2832 2833 2834 2835 2836 2837 2838
	return r;
}

static int dm_pr_clear(struct block_device *bdev, u64 key)
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2839
	int r;
2840

2841
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2842 2843 2844 2845 2846 2847 2848 2849 2850
	if (r < 0)
		return r;

	ops = bdev->bd_disk->fops->pr_ops;
	if (ops && ops->pr_clear)
		r = ops->pr_clear(bdev, key);
	else
		r = -EOPNOTSUPP;

2851
	bdput(bdev);
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862
	return r;
}

static const struct pr_ops dm_pr_ops = {
	.pr_register	= dm_pr_register,
	.pr_reserve	= dm_pr_reserve,
	.pr_release	= dm_pr_release,
	.pr_preempt	= dm_pr_preempt,
	.pr_clear	= dm_pr_clear,
};

2863
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2864 2865
	.open = dm_blk_open,
	.release = dm_blk_close,
2866
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2867
	.getgeo = dm_blk_getgeo,
2868
	.pr_ops = &dm_pr_ops,
L
Linus Torvalds 已提交
2869 2870 2871
	.owner = THIS_MODULE
};

2872 2873 2874 2875
static const struct dax_operations dm_dax_ops = {
	.direct_access = dm_dax_direct_access,
};

L
Linus Torvalds 已提交
2876 2877 2878 2879 2880 2881 2882 2883
/*
 * 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");
2884

2885 2886 2887
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

2888 2889 2890
module_param(dm_numa_node, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dm_numa_node, "NUMA node for DM device memory allocations");

L
Linus Torvalds 已提交
2891 2892 2893
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");