dm.c 67.7 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
 *
 * 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 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
/*
 * The zone descriptors obtained with a zone report indicate
 * zone positions within the target device. The zone descriptors
 * must be remapped to match their position within the dm device.
 * A target may call dm_remap_zone_report after completion of a
 * REQ_OP_ZONE_REPORT bio to remap the zone descriptors obtained
 * from the target device mapping to the dm device.
 */
void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, sector_t start)
{
#ifdef CONFIG_BLK_DEV_ZONED
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
	struct bio *report_bio = tio->io->bio;
	struct blk_zone_report_hdr *hdr = NULL;
	struct blk_zone *zone;
	unsigned int nr_rep = 0;
	unsigned int ofst;
	struct bio_vec bvec;
	struct bvec_iter iter;
	void *addr;

	if (bio->bi_status)
		return;

	/*
	 * Remap the start sector of the reported zones. For sequential zones,
	 * also remap the write pointer position.
	 */
	bio_for_each_segment(bvec, report_bio, iter) {
		addr = kmap_atomic(bvec.bv_page);

		/* Remember the report header in the first page */
		if (!hdr) {
			hdr = addr;
			ofst = sizeof(struct blk_zone_report_hdr);
		} else
			ofst = 0;

		/* Set zones start sector */
		while (hdr->nr_zones && ofst < bvec.bv_len) {
			zone = addr + ofst;
			if (zone->start >= start + ti->len) {
				hdr->nr_zones = 0;
				break;
			}
			zone->start = zone->start + ti->begin - start;
			if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) {
				if (zone->cond == BLK_ZONE_COND_FULL)
					zone->wp = zone->start + zone->len;
				else if (zone->cond == BLK_ZONE_COND_EMPTY)
					zone->wp = zone->start;
				else
					zone->wp = zone->wp + ti->begin - start;
			}
			ofst += sizeof(struct blk_zone);
			hdr->nr_zones--;
			nr_rep++;
		}

		if (addr != hdr)
			kunmap_atomic(addr);

		if (!hdr->nr_zones)
			break;
	}

	if (hdr) {
		hdr->nr_zones = nr_rep;
		kunmap_atomic(hdr);
	}

	bio_advance(report_bio, report_bio->bi_iter.bi_size);

#else /* !CONFIG_BLK_DEV_ZONED */
	bio->bi_status = BLK_STS_NOTSUPP;
#endif
}
EXPORT_SYMBOL_GPL(dm_remap_zone_report);

1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
/*
 * 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;
1108
	int i;
1109 1110 1111 1112 1113 1114

	INIT_LIST_HEAD(&o->cb.list);

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

1115 1116 1117 1118 1119 1120
	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;
1121 1122
			if (unlikely(!bs) || bs == fs_bio_set ||
			    !bs->rescue_workqueue) {
1123 1124 1125 1126 1127 1128 1129 1130
				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);
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
		}
	}
}

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
1148
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1149 1150
{
	int r;
1151
	sector_t sector;
1152
	struct dm_offload o;
1153
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1154
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1155 1156 1157 1158 1159 1160 1161 1162 1163

	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);
1164
	sector = clone->bi_iter.bi_sector;
1165 1166

	dm_offload_start(&o);
M
Mikulas Patocka 已提交
1167
	r = ti->type->map(ti, clone);
1168 1169
	dm_offload_end(&o);

1170 1171 1172 1173
	switch (r) {
	case DM_MAPIO_SUBMITTED:
		break;
	case DM_MAPIO_REMAPPED:
L
Linus Torvalds 已提交
1174
		/* the bio has been remapped so dispatch it */
1175 1176
		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
1177
		generic_make_request(clone);
1178 1179
		break;
	case DM_MAPIO_KILL:
1180 1181 1182
		dec_pending(tio->io, BLK_STS_IOERR);
		free_tio(tio);
		break;
1183
	case DM_MAPIO_REQUEUE:
1184
		dec_pending(tio->io, BLK_STS_DM_REQUEUE);
1185
		free_tio(tio);
1186 1187
		break;
	default:
1188 1189
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196 1197 1198
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1199
	unsigned sector_count;
L
Linus Torvalds 已提交
1200 1201
};

1202
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1203
{
1204 1205
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1206 1207 1208 1209 1210
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1211 1212
static int clone_bio(struct dm_target_io *tio, struct bio *bio,
		     sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1213
{
1214
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1215

1216 1217
	__bio_clone_fast(clone, bio);

1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
	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);
1230 1231 1232
		if (r < 0)
			return r;
	}
A
Alasdair G Kergon 已提交
1233

1234 1235
	if (bio_op(bio) != REQ_OP_ZONE_REPORT)
		bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
1236 1237
	clone->bi_iter.bi_size = to_bytes(len);

1238
	if (unlikely(bio_integrity(bio) != NULL))
1239
		bio_integrity_trim(clone, 0, len);
1240 1241

	return 0;
L
Linus Torvalds 已提交
1242 1243
}

1244
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1245
				      struct dm_target *ti,
1246
				      unsigned target_bio_nr)
1247
{
1248 1249 1250
	struct dm_target_io *tio;
	struct bio *clone;

1251
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1252
	tio = container_of(clone, struct dm_target_io, clone);
1253 1254 1255

	tio->io = ci->io;
	tio->ti = ti;
1256
	tio->target_bio_nr = target_bio_nr;
1257 1258 1259 1260

	return tio;
}

1261 1262
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1263
				       unsigned target_bio_nr, unsigned *len)
1264
{
1265
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1266
	struct bio *clone = &tio->clone;
1267

1268 1269
	tio->len_ptr = len;

1270
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1271
	if (len)
1272
		bio_setup_sector(clone, ci->sector, *len);
1273

A
Alasdair G Kergon 已提交
1274
	__map_bio(tio);
1275 1276
}

1277
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1278
				  unsigned num_bios, unsigned *len)
1279
{
1280
	unsigned target_bio_nr;
1281

1282
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1283
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1284 1285
}

1286
static int __send_empty_flush(struct clone_info *ci)
1287
{
1288
	unsigned target_nr = 0;
1289 1290
	struct dm_target *ti;

1291
	BUG_ON(bio_has_data(ci->bio));
1292
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1293
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1294 1295 1296 1297

	return 0;
}

1298
static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1299
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1300
{
1301
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1302
	struct dm_target_io *tio;
1303 1304
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
1305
	int r = 0;
M
Mike Snitzer 已提交
1306

1307 1308 1309 1310 1311
	/*
	 * 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
1312

1313
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1314
		tio = alloc_tio(ci, ti, target_bio_nr);
1315
		tio->len_ptr = len;
1316
		r = clone_bio(tio, bio, sector, *len);
1317
		if (r < 0) {
1318
			free_tio(tio);
1319
			break;
1320
		}
1321 1322
		__map_bio(tio);
	}
1323 1324

	return r;
M
Mike Snitzer 已提交
1325 1326
}

1327
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1328

1329
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1330
{
1331
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1332 1333
}

1334
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1335
{
1336
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1337 1338
}

1339 1340 1341 1342 1343
static unsigned get_num_write_zeroes_bios(struct dm_target *ti)
{
	return ti->num_write_zeroes_bios;
}

M
Mike Snitzer 已提交
1344
typedef bool (*is_split_required_fn)(struct dm_target *ti);
1345

M
Mike Snitzer 已提交
1346 1347
static bool is_split_required_for_discard(struct dm_target *ti)
{
1348
	return ti->split_discard_bios;
1349 1350
}

1351 1352 1353
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
1354
{
M
Mike Snitzer 已提交
1355
	struct dm_target *ti;
1356
	unsigned len;
1357
	unsigned num_bios;
1358

1359 1360 1361 1362
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
1363

M
Mike Snitzer 已提交
1364
		/*
M
Mike Snitzer 已提交
1365 1366
		 * 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
1367
		 * reconfiguration might also have changed that since the
1368
		 * check was performed.
M
Mike Snitzer 已提交
1369
		 */
1370 1371
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1372
			return -EOPNOTSUPP;
1373

M
Mike Snitzer 已提交
1374
		if (is_split_required && !is_split_required(ti))
1375
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1376
		else
1377
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1378

1379
		__send_duplicate_bios(ci, ti, num_bios, &len);
1380

1381 1382
		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1383 1384

	return 0;
1385 1386
}

1387
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1388
{
1389 1390
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1391
}
1392

1393
static int __send_write_same(struct clone_info *ci)
1394
{
1395
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
1396 1397
}

1398 1399 1400 1401 1402
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
1403 1404 1405
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1406
static int __split_and_process_non_flush(struct clone_info *ci)
1407
{
1408
	struct bio *bio = ci->bio;
1409
	struct dm_target *ti;
1410
	unsigned len;
1411
	int r;
1412

M
Mike Christie 已提交
1413
	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
1414
		return __send_discard(ci);
M
Mike Christie 已提交
1415
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME))
1416
		return __send_write_same(ci);
1417 1418
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES))
		return __send_write_zeroes(ci);
1419

1420 1421 1422 1423
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1424 1425 1426 1427 1428
	if (bio_op(bio) == REQ_OP_ZONE_REPORT)
		len = ci->sector_count;
	else
		len = min_t(sector_t, max_io_len(ci->sector, ti),
			    ci->sector_count);
1429

1430 1431 1432
	r = __clone_and_map_data_bio(ci, ti, ci->sector, &len);
	if (r < 0)
		return r;
1433

1434 1435
	ci->sector += len;
	ci->sector_count -= len;
1436

1437
	return 0;
1438 1439
}

L
Linus Torvalds 已提交
1440
/*
1441
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1442
 */
M
Mikulas Patocka 已提交
1443 1444
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
1445
{
L
Linus Torvalds 已提交
1446
	struct clone_info ci;
1447
	int error = 0;
L
Linus Torvalds 已提交
1448

M
Mikulas Patocka 已提交
1449
	if (unlikely(!map)) {
1450
		bio_io_error(bio);
1451 1452
		return;
	}
1453

M
Mikulas Patocka 已提交
1454
	ci.map = map;
L
Linus Torvalds 已提交
1455 1456
	ci.md = md;
	ci.io = alloc_io(md);
1457
	ci.io->status = 0;
L
Linus Torvalds 已提交
1458 1459 1460
	atomic_set(&ci.io->io_count, 1);
	ci.io->bio = bio;
	ci.io->md = md;
1461
	spin_lock_init(&ci.io->endio_lock);
1462
	ci.sector = bio->bi_iter.bi_sector;
1463

1464
	start_io_acct(ci.io);
1465

J
Jens Axboe 已提交
1466
	if (bio->bi_opf & REQ_PREFLUSH) {
1467 1468
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1469
		error = __send_empty_flush(&ci);
1470
		/* dec_pending submits any data associated with flush */
1471 1472 1473 1474
	} else if (bio_op(bio) == REQ_OP_ZONE_RESET) {
		ci.bio = bio;
		ci.sector_count = 0;
		error = __split_and_process_non_flush(&ci);
1475
	} else {
1476
		ci.bio = bio;
1477
		ci.sector_count = bio_sectors(bio);
1478
		while (ci.sector_count && !error)
1479
			error = __split_and_process_non_flush(&ci);
1480
	}
1481

L
Linus Torvalds 已提交
1482
	/* drop the extra reference count */
1483
	dec_pending(ci.io, error);
1484
}
L
Linus Torvalds 已提交
1485 1486 1487
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/
1488

1489
/*
L
Linus Torvalds 已提交
1490 1491
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
1492
 */
1493
static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio)
1494
{
1495
	int rw = bio_data_dir(bio);
1496
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1497 1498
	int srcu_idx;
	struct dm_table *map;
1499

M
Mikulas Patocka 已提交
1500
	map = dm_get_live_table(md, &srcu_idx);
1501

1502
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1503

1504 1505
	/* 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
1506
		dm_put_live_table(md, srcu_idx);
1507

J
Jens Axboe 已提交
1508
		if (!(bio->bi_opf & REQ_RAHEAD))
1509 1510
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1511
			bio_io_error(bio);
1512
		return BLK_QC_T_NONE;
1513
	}
L
Linus Torvalds 已提交
1514

M
Mikulas Patocka 已提交
1515 1516
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1517
	return BLK_QC_T_NONE;
1518 1519
}

L
Linus Torvalds 已提交
1520 1521
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1522 1523 1524
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1525

1526
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mike Snitzer 已提交
1527
		if (dm_request_based(md)) {
1528
			/*
M
Mike Snitzer 已提交
1529 1530
			 * With request-based DM we only need to check the
			 * top-level queue for congestion.
1531
			 */
1532
			r = md->queue->backing_dev_info->wb.state & bdi_bits;
M
Mike Snitzer 已提交
1533 1534 1535
		} else {
			map = dm_get_live_table_fast(md);
			if (map)
1536
				r = dm_table_any_congested(map, bdi_bits);
M
Mike Snitzer 已提交
1537
			dm_put_live_table_fast(md);
1538 1539 1540
		}
	}

L
Linus Torvalds 已提交
1541 1542 1543 1544 1545 1546
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1547
static void free_minor(int minor)
L
Linus Torvalds 已提交
1548
{
1549
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1550
	idr_remove(&_minor_idr, minor);
1551
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1552 1553 1554 1555 1556
}

/*
 * See if the device with a specific minor # is free.
 */
1557
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1558
{
T
Tejun Heo 已提交
1559
	int r;
L
Linus Torvalds 已提交
1560 1561 1562 1563

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

T
Tejun Heo 已提交
1564
	idr_preload(GFP_KERNEL);
1565
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1566

T
Tejun Heo 已提交
1567
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1568

1569
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1570 1571 1572 1573
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1574 1575
}

1576
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1577
{
T
Tejun Heo 已提交
1578
	int r;
J
Jeff Mahoney 已提交
1579

T
Tejun Heo 已提交
1580
	idr_preload(GFP_KERNEL);
1581
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1582

T
Tejun Heo 已提交
1583
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1584

1585
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1586 1587 1588 1589 1590
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1591 1592
}

1593
static const struct block_device_operations dm_blk_dops;
1594
static const struct dax_operations dm_dax_ops;
L
Linus Torvalds 已提交
1595

1596 1597
static void dm_wq_work(struct work_struct *work);

1598
void dm_init_md_queue(struct mapped_device *md)
1599 1600 1601
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
1602
	 * devices.  The type of this dm device may not have been decided yet.
1603 1604 1605 1606 1607 1608 1609
	 * 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);
1610 1611 1612 1613 1614 1615

	/*
	 * 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;
1616
	md->queue->backing_dev_info->congested_data = md;
1617
}
1618

1619
void dm_init_normal_md_queue(struct mapped_device *md)
1620
{
1621
	md->use_blk_mq = false;
1622 1623 1624 1625 1626
	dm_init_md_queue(md);

	/*
	 * Initialize aspects of queue that aren't relevant for blk-mq
	 */
1627
	md->queue->backing_dev_info->congested_fn = dm_any_congested;
1628 1629 1630
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

1631 1632 1633 1634 1635 1636
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);
1637
	mempool_destroy(md->io_pool);
1638 1639 1640
	if (md->bs)
		bioset_free(md->bs);

1641 1642 1643 1644 1645 1646
	if (md->dax_dev) {
		kill_dax(md->dax_dev);
		put_dax(md->dax_dev);
		md->dax_dev = NULL;
	}

1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
	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);

1658 1659
	cleanup_srcu_struct(&md->io_barrier);

1660 1661 1662 1663
	if (md->bdev) {
		bdput(md->bdev);
		md->bdev = NULL;
	}
1664 1665

	dm_mq_cleanup_mapped_device(md);
1666 1667
}

L
Linus Torvalds 已提交
1668 1669 1670
/*
 * Allocate and initialise a blank device with a given minor.
 */
1671
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1672
{
1673
	int r, numa_node_id = dm_get_numa_node();
1674
	struct dax_device *dax_dev;
1675
	struct mapped_device *md;
1676
	void *old_md;
L
Linus Torvalds 已提交
1677

1678
	md = kzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1679 1680 1681 1682 1683
	if (!md) {
		DMWARN("unable to allocate device, out of memory.");
		return NULL;
	}

1684
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1685
		goto bad_module_get;
1686

L
Linus Torvalds 已提交
1687
	/* get a minor number for the dev */
1688
	if (minor == DM_ANY_MINOR)
1689
		r = next_free_minor(&minor);
1690
	else
1691
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1692
	if (r < 0)
M
Milan Broz 已提交
1693
		goto bad_minor;
L
Linus Torvalds 已提交
1694

M
Mikulas Patocka 已提交
1695 1696 1697 1698
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

1699
	md->numa_node_id = numa_node_id;
1700
	md->use_blk_mq = dm_use_blk_mq_default();
1701
	md->init_tio_pdu = false;
1702
	md->type = DM_TYPE_NONE;
1703
	mutex_init(&md->suspend_lock);
1704
	mutex_init(&md->type_lock);
1705
	mutex_init(&md->table_devices_lock);
1706
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1707
	atomic_set(&md->holders, 1);
1708
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1709
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1710 1711
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
1712
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
1713
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1714

1715
	md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1716
	if (!md->queue)
1717
		goto bad;
L
Linus Torvalds 已提交
1718

1719
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1720

1721
	md->disk = alloc_disk_node(1, numa_node_id);
L
Linus Torvalds 已提交
1722
	if (!md->disk)
1723
		goto bad;
L
Linus Torvalds 已提交
1724

1725 1726
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1727
	init_waitqueue_head(&md->wait);
1728
	INIT_WORK(&md->work, dm_wq_work);
1729
	init_waitqueue_head(&md->eventq);
1730
	init_completion(&md->kobj_holder.completion);
1731
	md->kworker_task = NULL;
1732

L
Linus Torvalds 已提交
1733 1734 1735 1736 1737 1738
	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);
1739 1740 1741 1742 1743 1744

	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
1745
	add_disk(md->disk);
M
Mike Anderson 已提交
1746
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1747

T
Tejun Heo 已提交
1748
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
1749
	if (!md->wq)
1750
		goto bad;
1751

M
Mikulas Patocka 已提交
1752 1753
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
1754
		goto bad;
M
Mikulas Patocka 已提交
1755

1756
	bio_init(&md->flush_bio, NULL, 0);
1757
	md->flush_bio.bi_bdev = md->bdev;
J
Jan Kara 已提交
1758
	md->flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC;
1759

M
Mikulas Patocka 已提交
1760 1761
	dm_stats_init(&md->stats);

1762
	/* Populate the mapping, nobody knows we exist yet */
1763
	spin_lock(&_minor_lock);
1764
	old_md = idr_replace(&_minor_idr, md, minor);
1765
	spin_unlock(&_minor_lock);
1766 1767 1768

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1769 1770
	return md;

1771 1772
bad:
	cleanup_mapped_device(md);
M
Mikulas Patocka 已提交
1773
bad_io_barrier:
L
Linus Torvalds 已提交
1774
	free_minor(minor);
M
Milan Broz 已提交
1775
bad_minor:
1776
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1777
bad_module_get:
L
Linus Torvalds 已提交
1778 1779 1780 1781
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1782 1783
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1784 1785
static void free_dev(struct mapped_device *md)
{
1786
	int minor = MINOR(disk_devt(md->disk));
1787

M
Mikulas Patocka 已提交
1788
	unlock_fs(md);
1789

1790
	cleanup_mapped_device(md);
1791

1792
	free_table_devices(&md->table_devices);
1793 1794 1795
	dm_stats_cleanup(&md->stats);
	free_minor(minor);

1796
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1797 1798 1799
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1800 1801
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1802
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1803

1804 1805
	if (md->bs) {
		/* The md already has necessary mempools. */
1806
		if (dm_table_bio_based(t)) {
1807 1808 1809 1810 1811 1812 1813 1814
			/*
			 * 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;
		}
1815 1816 1817 1818 1819 1820 1821 1822 1823
		/*
		 * 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
1824
	}
K
Kiyoshi Ueda 已提交
1825

1826
	BUG_ON(!p || md->io_pool || md->bs);
1827

K
Kiyoshi Ueda 已提交
1828 1829 1830 1831
	md->io_pool = p->io_pool;
	p->io_pool = NULL;
	md->bs = p->bs;
	p->bs = NULL;
1832

K
Kiyoshi Ueda 已提交
1833
out:
1834
	/* mempool bind completed, no longer need any mempools in the table */
K
Kiyoshi Ueda 已提交
1835 1836 1837
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
1838 1839 1840 1841 1842
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
1843 1844
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
1845 1846
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
1847 1848 1849 1850
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

1851
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
1852

L
Linus Torvalds 已提交
1853
	atomic_inc(&md->event_nr);
1854
	atomic_inc(&dm_global_event_nr);
L
Linus Torvalds 已提交
1855
	wake_up(&md->eventq);
1856
	wake_up(&dm_global_eventq);
L
Linus Torvalds 已提交
1857 1858
}

1859 1860 1861
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
1862
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
1863
{
1864 1865
	lockdep_assert_held(&md->suspend_lock);

1866
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
1867

1868
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
1869 1870
}

1871 1872 1873 1874 1875
/*
 * 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
1876
{
1877
	struct dm_table *old_map;
1878
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
1879 1880
	sector_t size;

1881 1882
	lockdep_assert_held(&md->suspend_lock);

L
Linus Torvalds 已提交
1883
	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
1884 1885 1886 1887

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
1888
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
1889 1890
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
1891
	__set_size(md, size);
1892

1893 1894
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
1895 1896 1897 1898 1899 1900 1901
	/*
	 * 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
1902
	if (dm_table_request_based(t)) {
1903
		dm_stop_queue(q);
M
Mike Snitzer 已提交
1904 1905 1906 1907 1908 1909 1910
		/*
		 * 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
1911 1912 1913

	__bind_mempools(md, t);

1914
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
1915
	rcu_assign_pointer(md->map, (void *)t);
1916 1917
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

1918
	dm_table_set_restrictions(t, q, limits);
1919 1920
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
1921

1922
	return old_map;
L
Linus Torvalds 已提交
1923 1924
}

1925 1926 1927 1928
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
1929
{
1930
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
1931 1932

	if (!map)
1933
		return NULL;
L
Linus Torvalds 已提交
1934 1935

	dm_table_event_callback(map, NULL, NULL);
1936
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
1937
	dm_sync_table(md);
1938 1939

	return map;
L
Linus Torvalds 已提交
1940 1941 1942 1943 1944
}

/*
 * Constructor for a new device.
 */
1945
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
1946 1947 1948
{
	struct mapped_device *md;

1949
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
1950 1951 1952
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
1953 1954
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
1955 1956 1957 1958
	*result = md;
	return 0;
}

1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
/*
 * 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);
}

1973
void dm_set_md_type(struct mapped_device *md, enum dm_queue_mode type)
1974
{
1975
	BUG_ON(!mutex_is_locked(&md->type_lock));
1976 1977 1978
	md->type = type;
}

1979
enum dm_queue_mode dm_get_md_type(struct mapped_device *md)
1980 1981 1982 1983
{
	return md->type;
}

1984 1985 1986 1987 1988
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
/*
 * 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);

2000 2001 2002
/*
 * Setup the DM device's queue based on md's type
 */
2003
int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t)
2004
{
2005
	int r;
2006
	enum dm_queue_mode type = dm_get_md_type(md);
2007

2008
	switch (type) {
2009
	case DM_TYPE_REQUEST_BASED:
2010
		r = dm_old_init_request_queue(md, t);
2011
		if (r) {
2012
			DMERR("Cannot initialize queue for request-based mapped device");
2013
			return r;
2014
		}
2015 2016
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
2017
		r = dm_mq_init_request_queue(md, t);
2018
		if (r) {
2019
			DMERR("Cannot initialize queue for request-based dm-mq mapped device");
2020 2021 2022 2023
			return r;
		}
		break;
	case DM_TYPE_BIO_BASED:
2024
	case DM_TYPE_DAX_BIO_BASED:
2025
		dm_init_normal_md_queue(md);
2026
		blk_queue_make_request(md->queue, dm_make_request);
2027 2028 2029 2030 2031 2032
		/*
		 * 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;
2033 2034 2035

		if (type == DM_TYPE_DAX_BIO_BASED)
			queue_flag_set_unlocked(QUEUE_FLAG_DAX, md->queue);
2036
		break;
2037 2038 2039
	case DM_TYPE_NONE:
		WARN_ON_ONCE(true);
		break;
2040 2041 2042 2043 2044
	}

	return 0;
}

2045
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
2046 2047 2048 2049 2050 2051 2052
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2053
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2054 2055

	md = idr_find(&_minor_idr, minor);
2056 2057 2058 2059 2060 2061 2062 2063 2064
	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
2065
	}
L
Linus Torvalds 已提交
2066

J
Jeff Mahoney 已提交
2067
out:
2068
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2069

2070 2071
	return md;
}
A
Alasdair G Kergon 已提交
2072
EXPORT_SYMBOL_GPL(dm_get_md);
2073

A
Alasdair G Kergon 已提交
2074
void *dm_get_mdptr(struct mapped_device *md)
2075
{
A
Alasdair G Kergon 已提交
2076
	return md->interface_ptr;
L
Linus Torvalds 已提交
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
}

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);
2087
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2088 2089
}

2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
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);

2103 2104 2105 2106 2107 2108
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2109
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2110
{
2111
	struct request_queue *q = dm_get_md_queue(md);
M
Mike Anderson 已提交
2112
	struct dm_table *map;
M
Mikulas Patocka 已提交
2113
	int srcu_idx;
L
Linus Torvalds 已提交
2114

2115
	might_sleep();
J
Jeff Mahoney 已提交
2116

2117
	spin_lock(&_minor_lock);
2118 2119 2120
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);
2121

2122
	blk_set_queue_dying(q);
2123

2124
	if (dm_request_based(md) && md->kworker_task)
P
Petr Mladek 已提交
2125
		kthread_flush_worker(&md->kworker);
2126

2127 2128 2129 2130 2131
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
2132
	map = dm_get_live_table(md, &srcu_idx);
2133 2134 2135
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2136
	}
M
Mikulas Patocka 已提交
2137 2138
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);
2139
	mutex_unlock(&md->suspend_lock);
M
Mikulas Patocka 已提交
2140

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
	/*
	 * 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
2172
}
E
Edward Goggin 已提交
2173
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2174

2175
static int dm_wait_for_completion(struct mapped_device *md, long task_state)
2176 2177
{
	int r = 0;
2178
	DEFINE_WAIT(wait);
2179 2180

	while (1) {
2181
		prepare_to_wait(&md->wait, &wait, task_state);
2182

2183
		if (!md_in_flight(md))
2184 2185
			break;

2186
		if (signal_pending_state(task_state, current)) {
2187 2188 2189 2190 2191 2192
			r = -EINTR;
			break;
		}

		io_schedule();
	}
2193
	finish_wait(&md->wait, &wait);
2194

2195 2196 2197
	return r;
}

L
Linus Torvalds 已提交
2198 2199 2200
/*
 * Process the deferred bios
 */
2201
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2202
{
2203 2204
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2205
	struct bio *c;
M
Mikulas Patocka 已提交
2206 2207
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2208

M
Mikulas Patocka 已提交
2209
	map = dm_get_live_table(md, &srcu_idx);
2210

2211
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2212 2213 2214 2215
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2216
		if (!c)
A
Alasdair G Kergon 已提交
2217
			break;
2218

K
Kiyoshi Ueda 已提交
2219 2220
		if (dm_request_based(md))
			generic_make_request(c);
2221
		else
M
Mikulas Patocka 已提交
2222
			__split_and_process_bio(md, map, c);
2223
	}
M
Milan Broz 已提交
2224

M
Mikulas Patocka 已提交
2225
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2226 2227
}

2228
static void dm_queue_flush(struct mapped_device *md)
2229
{
2230
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2231
	smp_mb__after_atomic();
2232
	queue_work(md->wq, &md->work);
2233 2234
}

L
Linus Torvalds 已提交
2235
/*
2236
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2237
 */
2238
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2239
{
2240
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2241
	struct queue_limits limits;
2242
	int r;
L
Linus Torvalds 已提交
2243

2244
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2245 2246

	/* device must be suspended */
2247
	if (!dm_suspended_md(md))
2248
		goto out;
L
Linus Torvalds 已提交
2249

2250 2251 2252 2253 2254 2255 2256
	/*
	 * 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
2257
		live_map = dm_get_live_table_fast(md);
2258 2259
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2260
		dm_put_live_table_fast(md);
2261 2262
	}

2263 2264 2265 2266 2267 2268
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2269
	}
2270

2271
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2272

2273
out:
2274
	mutex_unlock(&md->suspend_lock);
2275
	return map;
L
Linus Torvalds 已提交
2276 2277 2278 2279 2280 2281
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2282
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2283
{
2284
	int r;
L
Linus Torvalds 已提交
2285 2286

	WARN_ON(md->frozen_sb);
2287

2288
	md->frozen_sb = freeze_bdev(md->bdev);
2289
	if (IS_ERR(md->frozen_sb)) {
2290
		r = PTR_ERR(md->frozen_sb);
2291 2292
		md->frozen_sb = NULL;
		return r;
2293 2294
	}

2295 2296
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2297 2298 2299
	return 0;
}

2300
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2301
{
2302 2303 2304
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2305
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2306
	md->frozen_sb = NULL;
2307
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2308 2309 2310
}

/*
2311 2312 2313 2314
 * @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
 *
2315 2316 2317
 * 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.
2318
 */
2319
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
2320
			unsigned suspend_flags, long task_state,
2321
			int dmf_suspended_flag)
L
Linus Torvalds 已提交
2322
{
2323 2324 2325
	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
2326

2327 2328
	lockdep_assert_held(&md->suspend_lock);

2329 2330 2331 2332 2333 2334
	/*
	 * 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);
2335 2336
	else
		pr_debug("%s: suspending with flush\n", dm_device_name(md));
2337

2338 2339 2340 2341
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
2342 2343
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2344
	/*
K
Kiyoshi Ueda 已提交
2345 2346 2347 2348
	 * 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
2349 2350 2351
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
2352 2353
		if (r) {
			dm_table_presuspend_undo_targets(map);
2354
			return r;
2355
		}
2356
	}
L
Linus Torvalds 已提交
2357 2358

	/*
2359 2360 2361 2362 2363 2364 2365
	 * 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
2366 2367 2368
	 * __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
2369
	 */
2370
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2371 2372
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2373

2374
	/*
2375 2376
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2377
	 */
2378
	if (dm_request_based(md)) {
2379
		dm_stop_queue(md->queue);
2380
		if (md->kworker_task)
P
Petr Mladek 已提交
2381
			kthread_flush_worker(&md->kworker);
2382
	}
2383

2384 2385
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2386
	/*
2387 2388 2389
	 * 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
2390
	 */
2391
	r = dm_wait_for_completion(md, task_state);
2392 2393
	if (!r)
		set_bit(dmf_suspended_flag, &md->flags);
L
Linus Torvalds 已提交
2394

2395
	if (noflush)
2396
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2397 2398
	if (map)
		synchronize_srcu(&md->io_barrier);
2399

L
Linus Torvalds 已提交
2400
	/* were we interrupted ? */
2401
	if (r < 0) {
2402
		dm_queue_flush(md);
M
Milan Broz 已提交
2403

2404
		if (dm_request_based(md))
2405
			dm_start_queue(md->queue);
2406

2407
		unlock_fs(md);
2408
		dm_table_presuspend_undo_targets(map);
2409
		/* pushback list is already flushed, so skip flush */
2410
	}
L
Linus Torvalds 已提交
2411

2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452
	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;
	}

2453
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2454

2455
	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED);
2456 2457
	if (r)
		goto out_unlock;
2458

2459 2460
	dm_table_postsuspend_targets(map);

2461
out_unlock:
2462
	mutex_unlock(&md->suspend_lock);
2463
	return r;
L
Linus Torvalds 已提交
2464 2465
}

2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481
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))
2482
		dm_start_queue(md->queue);
2483 2484 2485 2486 2487 2488

	unlock_fs(md);

	return 0;
}

L
Linus Torvalds 已提交
2489 2490
int dm_resume(struct mapped_device *md)
{
2491
	int r;
2492
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2493

2494
retry:
2495
	r = -EINVAL;
2496 2497
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

2498
	if (!dm_suspended_md(md))
2499 2500
		goto out;

2501 2502 2503 2504 2505 2506 2507 2508 2509
	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;
	}

2510
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2511
	if (!map || !dm_table_get_size(map))
2512
		goto out;
L
Linus Torvalds 已提交
2513

2514
	r = __dm_resume(md, map);
2515 2516
	if (r)
		goto out;
2517 2518

	clear_bit(DMF_SUSPENDED, &md->flags);
2519
out:
2520
	mutex_unlock(&md->suspend_lock);
2521

2522
	return r;
L
Linus Torvalds 已提交
2523 2524
}

M
Mikulas Patocka 已提交
2525 2526 2527 2528 2529 2530
/*
 * 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.
 */

2531
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
2532
{
2533 2534
	struct dm_table *map = NULL;

2535 2536
	lockdep_assert_held(&md->suspend_lock);

2537
	if (md->internal_suspend_count++)
2538 2539 2540 2541 2542 2543 2544
		return; /* nested internal suspend */

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

2545
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2546 2547 2548 2549 2550 2551 2552

	/*
	 * 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.
	 */
2553 2554
	(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE,
			    DMF_SUSPENDED_INTERNALLY);
2555 2556 2557 2558 2559 2560

	dm_table_postsuspend_targets(map);
}

static void __dm_internal_resume(struct mapped_device *md)
{
2561 2562 2563
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
2564 2565
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
2566
	if (dm_suspended_md(md))
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
		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
2606 2607 2608 2609 2610 2611 2612
		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);
}
2613
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
2614

2615
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
2616
{
2617
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
2618 2619 2620 2621 2622 2623 2624
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
2625
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
2626

L
Linus Torvalds 已提交
2627 2628 2629
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2630
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2631
		       unsigned cookie)
2632
{
M
Milan Broz 已提交
2633 2634 2635 2636
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2637
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2638 2639 2640
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2641 2642
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2643
	}
2644 2645
}

M
Mike Anderson 已提交
2646 2647 2648 2649 2650
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661
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
2662 2663 2664 2665 2666 2667 2668 2669 2670
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
2671 2672 2673 2674 2675 2676 2677 2678
/*
 * 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;
}
2679
EXPORT_SYMBOL_GPL(dm_disk);
L
Linus Torvalds 已提交
2680

M
Milan Broz 已提交
2681 2682
struct kobject *dm_kobject(struct mapped_device *md)
{
2683
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
2684 2685 2686 2687 2688 2689
}

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

2690
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
2691

2692
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2693
	    dm_deleting_md(md))
2694 2695
		return NULL;

M
Milan Broz 已提交
2696 2697 2698 2699
	dm_get(md);
	return md;
}

2700
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2701 2702 2703 2704
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2705 2706 2707 2708 2709
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
2710 2711 2712 2713 2714
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

2715 2716
int dm_suspended(struct dm_target *ti)
{
2717
	return dm_suspended_md(dm_table_get_md(ti->table));
2718 2719 2720
}
EXPORT_SYMBOL_GPL(dm_suspended);

2721 2722
int dm_noflush_suspending(struct dm_target *ti)
{
2723
	return __noflush_suspending(dm_table_get_md(ti->table));
2724 2725 2726
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

2727
struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type,
2728
					    unsigned integrity, unsigned per_io_data_size)
K
Kiyoshi Ueda 已提交
2729
{
2730
	struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id);
2731
	unsigned int pool_size = 0;
J
Jun'ichi Nomura 已提交
2732
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2733 2734

	if (!pools)
2735
		return NULL;
K
Kiyoshi Ueda 已提交
2736

2737 2738
	switch (type) {
	case DM_TYPE_BIO_BASED:
2739
	case DM_TYPE_DAX_BIO_BASED:
2740
		pool_size = dm_get_reserved_bio_based_ios();
2741
		front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
2742 2743 2744 2745
	
		pools->io_pool = mempool_create_slab_pool(pool_size, _io_cache);
		if (!pools->io_pool)
			goto out;
2746 2747 2748
		break;
	case DM_TYPE_REQUEST_BASED:
	case DM_TYPE_MQ_REQUEST_BASED:
2749
		pool_size = dm_get_reserved_rq_based_ios();
2750
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
2751
		/* per_io_data_size is used for blk-mq pdu at queue allocation */
2752 2753 2754 2755 2756
		break;
	default:
		BUG();
	}

2757
	pools->bs = bioset_create(pool_size, front_pad, BIOSET_NEED_RESCUER);
K
Kiyoshi Ueda 已提交
2758
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2759
		goto out;
K
Kiyoshi Ueda 已提交
2760

2761
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2762
		goto out;
2763

K
Kiyoshi Ueda 已提交
2764
	return pools;
2765 2766 2767

out:
	dm_free_md_mempools(pools);
2768

2769
	return NULL;
K
Kiyoshi Ueda 已提交
2770 2771 2772 2773 2774 2775 2776
}

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

2777
	mempool_destroy(pools->io_pool);
2778

K
Kiyoshi Ueda 已提交
2779 2780 2781 2782 2783 2784
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

2785 2786 2787 2788 2789 2790 2791 2792 2793
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)
2794 2795
{
	struct mapped_device *md = bdev->bd_disk->private_data;
2796 2797 2798
	struct dm_table *table;
	struct dm_target *ti;
	int ret = -ENOTTY, srcu_idx;
2799

2800 2801 2802
	table = dm_get_live_table(md, &srcu_idx);
	if (!table || !dm_table_get_size(table))
		goto out;
2803

2804 2805 2806 2807
	/* 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);
2808

2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854
	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;
2855 2856 2857
}

static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2858
			 u32 flags)
2859 2860 2861 2862
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2863
	int r;
2864

2865
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2866 2867 2868 2869 2870 2871 2872 2873 2874
	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;

2875
	bdput(bdev);
2876 2877 2878 2879 2880 2881 2882 2883
	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;
2884
	int r;
2885

2886
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2887 2888 2889 2890 2891 2892 2893 2894 2895
	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;

2896
	bdput(bdev);
2897 2898 2899 2900
	return r;
}

static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2901
			 enum pr_type type, bool abort)
2902 2903 2904 2905
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2906
	int r;
2907

2908
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2909 2910 2911 2912 2913 2914 2915 2916 2917
	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;

2918
	bdput(bdev);
2919 2920 2921 2922 2923 2924 2925 2926
	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;
2927
	int r;
2928

2929
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2930 2931 2932 2933 2934 2935 2936 2937 2938
	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;

2939
	bdput(bdev);
2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
	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,
};

2951
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2952 2953
	.open = dm_blk_open,
	.release = dm_blk_close,
2954
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2955
	.getgeo = dm_blk_getgeo,
2956
	.pr_ops = &dm_pr_ops,
L
Linus Torvalds 已提交
2957 2958 2959
	.owner = THIS_MODULE
};

2960 2961 2962 2963
static const struct dax_operations dm_dax_ops = {
	.direct_access = dm_dax_direct_access,
};

L
Linus Torvalds 已提交
2964 2965 2966 2967 2968 2969 2970 2971
/*
 * 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");
2972

2973 2974 2975
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

2976 2977 2978
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
2979 2980 2981
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");