dm.c 65.1 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;

L
Linus Torvalds 已提交
61 62 63 64 65 66 67
/*
 * One of these is allocated per bio.
 */
struct dm_io {
	struct mapped_device *md;
	int error;
	atomic_t io_count;
R
Richard Kennedy 已提交
68
	struct bio *bio;
69
	unsigned long start_time;
70
	spinlock_t endio_lock;
M
Mikulas Patocka 已提交
71
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
72 73
};

74 75
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
76 77 78
/*
 * Bits for the md->flags field.
 */
79
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
80
#define DMF_SUSPENDED 1
81
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
82
#define DMF_FREEING 3
83
#define DMF_DELETING 4
84
#define DMF_NOFLUSH_SUSPENDING 5
85 86
#define DMF_DEFERRED_REMOVE 6
#define DMF_SUSPENDED_INTERNALLY 7
L
Linus Torvalds 已提交
87

88 89
#define DM_NUMA_NODE NUMA_NO_NODE
static int dm_numa_node = DM_NUMA_NODE;
90

K
Kiyoshi Ueda 已提交
91 92 93 94 95 96 97 98
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
	struct bio_set *bs;
};

99 100 101 102 103 104
struct table_device {
	struct list_head list;
	atomic_t count;
	struct dm_dev dm_dev;
};

105
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
106
static struct kmem_cache *_rq_tio_cache;
107
static struct kmem_cache *_rq_cache;
108

109 110 111
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
112
#define RESERVED_BIO_BASED_IOS		16
113 114
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
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;
}

136 137
unsigned __dm_get_module_param(unsigned *module_param,
			       unsigned def, unsigned max)
138
{
139 140
	unsigned param = ACCESS_ONCE(*module_param);
	unsigned modified_param = 0;
141

142 143 144 145
	if (!param)
		modified_param = def;
	else if (param > max)
		modified_param = max;
146

147 148 149
	if (modified_param) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
150 151
	}

152
	return param;
153 154
}

155 156
unsigned dm_get_reserved_bio_based_ios(void)
{
157
	return __dm_get_module_param(&reserved_bio_based_ios,
158
				     RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS);
159 160 161
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

162 163 164 165 166 167
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
168 169
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
170
	int r = -ENOMEM;
L
Linus Torvalds 已提交
171 172

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
173
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
174
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
175
		return r;
L
Linus Torvalds 已提交
176

K
Kiyoshi Ueda 已提交
177 178
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
179
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
180

181
	_rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request),
182 183 184 185
				      __alignof__(struct request), 0, NULL);
	if (!_rq_cache)
		goto out_free_rq_tio_cache;

M
Mike Anderson 已提交
186
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
187
	if (r)
188
		goto out_free_rq_cache;
M
Mike Anderson 已提交
189

190 191 192 193 194 195
	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
196 197
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
198
	if (r < 0)
199
		goto out_free_workqueue;
L
Linus Torvalds 已提交
200 201 202 203 204

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
205

206 207
out_free_workqueue:
	destroy_workqueue(deferred_remove_workqueue);
K
Kiyoshi Ueda 已提交
208 209
out_uevent_exit:
	dm_uevent_exit();
210 211
out_free_rq_cache:
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
212 213
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
214 215 216 217
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
218 219 220 221
}

static void local_exit(void)
{
M
Mikulas Patocka 已提交
222
	flush_scheduled_work();
223
	destroy_workqueue(deferred_remove_workqueue);
M
Mikulas Patocka 已提交
224

225
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
226
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
227
	kmem_cache_destroy(_io_cache);
228
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
229
	dm_uevent_exit();
L
Linus Torvalds 已提交
230 231 232 233 234 235

	_major = 0;

	DMINFO("cleaned up");
}

236
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
237 238 239 240
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
241
	dm_io_init,
242
	dm_kcopyd_init,
L
Linus Torvalds 已提交
243
	dm_interface_init,
M
Mikulas Patocka 已提交
244
	dm_statistics_init,
L
Linus Torvalds 已提交
245 246
};

247
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
248 249 250 251
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
252
	dm_io_exit,
253
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
254
	dm_interface_exit,
M
Mikulas Patocka 已提交
255
	dm_statistics_exit,
L
Linus Torvalds 已提交
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
};

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]();
285 286 287 288 289

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
290 291 292 293 294
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
295 296 297 298 299
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
300
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
301 302 303
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
304 305
	spin_lock(&_minor_lock);

A
Al Viro 已提交
306
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
307 308 309
	if (!md)
		goto out;

310
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
311
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
312 313 314 315
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
316
	dm_get(md);
317
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
318 319 320 321
out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
322 323
}

324
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
325
{
326
	struct mapped_device *md;
327

328 329
	spin_lock(&_minor_lock);

330 331 332 333
	md = disk->private_data;
	if (WARN_ON(!md))
		goto out;

M
Mikulas Patocka 已提交
334 335
	if (atomic_dec_and_test(&md->open_count) &&
	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
336
		queue_work(deferred_remove_workqueue, &deferred_remove_work);
M
Mikulas Patocka 已提交
337

L
Linus Torvalds 已提交
338
	dm_put(md);
339
out:
340
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
341 342
}

343 344 345 346 347 348 349 350
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
351
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
352 353 354 355 356
{
	int r = 0;

	spin_lock(&_minor_lock);

M
Mikulas Patocka 已提交
357
	if (dm_open_count(md)) {
358
		r = -EBUSY;
M
Mikulas Patocka 已提交
359 360 361 362
		if (mark_deferred)
			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
		r = -EEXIST;
363 364 365 366 367 368 369 370
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

M
Mikulas Patocka 已提交
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391
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
392 393 394 395 396
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

397 398 399 400 401
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
	return md->queue;
}

M
Mikulas Patocka 已提交
402 403 404 405 406
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
407 408 409 410 411 412 413
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);
}

414 415 416
static int dm_grab_bdev_for_ioctl(struct mapped_device *md,
				  struct block_device **bdev,
				  fmode_t *mode)
417
{
418
	struct dm_target *tgt;
419
	struct dm_table *map;
420
	int srcu_idx, r;
421

422
retry:
C
Christoph Hellwig 已提交
423
	r = -ENOTTY;
424
	map = dm_get_live_table(md, &srcu_idx);
425 426 427 428 429 430 431
	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;

432 433
	tgt = dm_table_get_target(map, 0);
	if (!tgt->type->prepare_ioctl)
434
		goto out;
435

436
	if (dm_suspended_md(md)) {
437 438 439 440
		r = -EAGAIN;
		goto out;
	}

441
	r = tgt->type->prepare_ioctl(tgt, bdev, mode);
C
Christoph Hellwig 已提交
442 443
	if (r < 0)
		goto out;
444

445 446
	bdgrab(*bdev);
	dm_put_live_table(md, srcu_idx);
C
Christoph Hellwig 已提交
447
	return r;
448 449

out:
450
	dm_put_live_table(md, srcu_idx);
451
	if (r == -ENOTCONN && !fatal_signal_pending(current)) {
452 453 454
		msleep(10);
		goto retry;
	}
C
Christoph Hellwig 已提交
455 456 457 458 459 460 461
	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;
462
	int r;
C
Christoph Hellwig 已提交
463

464
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
C
Christoph Hellwig 已提交
465 466
	if (r < 0)
		return r;
467

C
Christoph Hellwig 已提交
468 469
	if (r > 0) {
		/*
470 471
		 * Target determined this ioctl is being issued against a
		 * subset of the parent bdev; require extra privileges.
C
Christoph Hellwig 已提交
472
		 */
473 474 475 476 477
		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
478
			goto out;
479
		}
C
Christoph Hellwig 已提交
480
	}
481

482
	r =  __blkdev_driver_ioctl(bdev, mode, cmd, arg);
C
Christoph Hellwig 已提交
483
out:
484
	bdput(bdev);
485 486 487
	return r;
}

A
Alasdair G Kergon 已提交
488
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
489 490 491 492
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
493
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
494 495 496 497
{
	mempool_free(io, md->io_pool);
}

498
static void free_tio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
499
{
500
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
501 502
}

503
int md_in_flight(struct mapped_device *md)
K
Kiyoshi Ueda 已提交
504 505 506 507 508
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

509 510 511
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
512
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
513
	int cpu;
M
Mikulas Patocka 已提交
514
	int rw = bio_data_dir(bio);
515 516 517

	io->start_time = jiffies;

T
Tejun Heo 已提交
518 519 520
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
521 522
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
523 524

	if (unlikely(dm_stats_used(&md->stats)))
525 526 527
		dm_stats_account_io(&md->stats, bio_data_dir(bio),
				    bio->bi_iter.bi_sector, bio_sectors(bio),
				    false, 0, &io->stats_aux);
528 529
}

530
static void end_io_acct(struct dm_io *io)
531 532 533 534
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
535
	int pending;
536 537
	int rw = bio_data_dir(bio);

538
	generic_end_io_acct(rw, &dm_disk(md)->part0, io->start_time);
539

M
Mikulas Patocka 已提交
540
	if (unlikely(dm_stats_used(&md->stats)))
541 542 543
		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
544

545 546
	/*
	 * After this is decremented the bio must not be touched if it is
547
	 * a flush.
548
	 */
549 550
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
551
	pending += atomic_read(&md->pending[rw^0x1]);
552

553 554 555
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
556 557
}

L
Linus Torvalds 已提交
558 559 560
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
561
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
562
{
563
	unsigned long flags;
L
Linus Torvalds 已提交
564

565
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
566
	bio_list_add(&md->deferred, bio);
567
	spin_unlock_irqrestore(&md->deferred_lock, flags);
568
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
569 570 571 572 573
}

/*
 * 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
574
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
575
 */
M
Mikulas Patocka 已提交
576
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
577
{
M
Mikulas Patocka 已提交
578 579 580 581
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
582

M
Mikulas Patocka 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
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
603

M
Mikulas Patocka 已提交
604 605 606
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
607 608
}

609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
/*
 * 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;
633
	td->dm_dev.dax_dev = dax_get_by_host(bdev->bd_disk->disk_name);
634 635 636 637 638 639 640 641 642 643 644 645 646
	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);
647
	put_dax(td->dm_dev.dax_dev);
648
	td->dm_dev.bdev = NULL;
649
	td->dm_dev.dax_dev = NULL;
650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670
}

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) {
671
		td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id);
672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
		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
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752
/*
 * 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
753 754 755 756 757 758 759 760 761
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

762 763 764 765 766
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
767 768 769 770
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
771
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
772
{
773
	unsigned long flags;
774 775 776
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
777 778

	/* Push-back supersedes any I/O errors */
779 780 781 782 783 784
	if (unlikely(error)) {
		spin_lock_irqsave(&io->endio_lock, flags);
		if (!(io->error > 0 && __noflush_suspending(md)))
			io->error = error;
		spin_unlock_irqrestore(&io->endio_lock, flags);
	}
L
Linus Torvalds 已提交
785 786

	if (atomic_dec_and_test(&io->io_count)) {
787 788 789 790
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
791
			spin_lock_irqsave(&md->deferred_lock, flags);
792 793 794
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
795 796
				/* noflush suspend was interrupted. */
				io->error = -EIO;
797
			spin_unlock_irqrestore(&md->deferred_lock, flags);
798 799
		}

800 801
		io_error = io->error;
		bio = io->bio;
802 803 804 805 806
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
807

J
Jens Axboe 已提交
808
		if ((bio->bi_opf & REQ_PREFLUSH) && bio->bi_iter.bi_size) {
809
			/*
810
			 * Preflush done for flush with data, reissue
811
			 * without REQ_PREFLUSH.
812
			 */
J
Jens Axboe 已提交
813
			bio->bi_opf &= ~REQ_PREFLUSH;
814
			queue_io(md, bio);
815
		} else {
816
			/* done with normal IO or empty flush */
817 818
			bio->bi_error = io_error;
			bio_endio(bio);
819
		}
L
Linus Torvalds 已提交
820 821 822
	}
}

823
void disable_write_same(struct mapped_device *md)
824 825 826 827 828 829 830
{
	struct queue_limits *limits = dm_get_queue_limits(md);

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

831 832 833 834 835 836 837 838
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;
}

839
static void clone_endio(struct bio *bio)
L
Linus Torvalds 已提交
840
{
841
	int error = bio->bi_error;
842
	int r = error;
M
Mikulas Patocka 已提交
843
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
844
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
845
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
846 847
	dm_endio_fn endio = tio->ti->type->end_io;

848
	if (unlikely(error == -EREMOTEIO)) {
849 850 851 852 853 854 855
		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);
	}
856

857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873
	if (endio) {
		r = endio(tio->ti, bio, &error);
		switch (r) {
		case DM_ENDIO_REQUEUE:
			error = DM_ENDIO_REQUEUE;
			/*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();
		}
	}

874
	free_tio(tio);
875
	dec_pending(io, error);
L
Linus Torvalds 已提交
876 877
}

878 879 880 881 882 883 884 885 886 887 888 889
/*
 * 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
890
{
891
	sector_t len = max_io_len_target_boundary(sector, ti);
892
	sector_t offset, max_len;
L
Linus Torvalds 已提交
893 894

	/*
895
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
896
	 */
897 898 899 900 901 902 903 904 905 906
	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
907 908 909 910 911
	}

	return len;
}

912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
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);

927 928
static struct dm_target *dm_dax_get_live_target(struct mapped_device *md,
		sector_t sector, int *srcu_idx)
929 930 931 932
{
	struct dm_table *map;
	struct dm_target *ti;

933
	map = dm_get_live_table(md, srcu_idx);
934
	if (!map)
935
		return NULL;
936 937 938

	ti = dm_table_find_target(map, sector);
	if (!dm_target_is_valid(ti))
939
		return NULL;
940

941 942
	return ti;
}
943

944 945 946 947 948 949 950 951
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;
952

953
	ti = dm_dax_get_live_target(md, sector, &srcu_idx);
954

955 956 957 958 959 960 961 962
	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);
963
	if (ti->type->direct_access)
964 965
		ret = ti->type->direct_access(ti, pgoff, nr_pages, kaddr, pfn);

966
 out:
967
	dm_put_live_table(md, srcu_idx);
968 969

	return ret;
970 971
}

972 973
/*
 * A target may call dm_accept_partial_bio only from the map routine.  It is
974
 * allowed for all bio types except REQ_PREFLUSH.
975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
 *
 * 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
1004
	BUG_ON(bio->bi_opf & REQ_PREFLUSH);
1005 1006 1007 1008 1009 1010 1011
	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);

1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
/*
 * 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;
1026
	int i;
1027 1028 1029 1030 1031 1032

	INIT_LIST_HEAD(&o->cb.list);

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

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
	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;
			if (unlikely(!bs) || bs == fs_bio_set) {
				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);
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
		}
	}
}

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
1065
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1066 1067
{
	int r;
1068
	sector_t sector;
1069
	struct dm_offload o;
1070
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1071
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080

	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);
1081
	sector = clone->bi_iter.bi_sector;
1082 1083

	dm_offload_start(&o);
M
Mikulas Patocka 已提交
1084
	r = ti->type->map(ti, clone);
1085 1086
	dm_offload_end(&o);

1087 1088 1089 1090
	switch (r) {
	case DM_MAPIO_SUBMITTED:
		break;
	case DM_MAPIO_REMAPPED:
L
Linus Torvalds 已提交
1091
		/* the bio has been remapped so dispatch it */
1092 1093
		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
1094
		generic_make_request(clone);
1095 1096 1097 1098 1099
		break;
	case DM_MAPIO_KILL:
		r = -EIO;
		/*FALLTHRU*/
	case DM_MAPIO_REQUEUE:
1100
		/* error the io and bail out, or requeue it if needed */
S
Stefan Bader 已提交
1101
		dec_pending(tio->io, r);
1102
		free_tio(tio);
1103 1104
		break;
	default:
1105 1106
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1107 1108 1109 1110 1111 1112 1113 1114 1115
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1116
	unsigned sector_count;
L
Linus Torvalds 已提交
1117 1118
};

1119
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1120
{
1121 1122
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1123 1124 1125 1126 1127
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1128 1129
static int clone_bio(struct dm_target_io *tio, struct bio *bio,
		     sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1130
{
1131
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1132

1133 1134
	__bio_clone_fast(clone, bio);

1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
	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);
1147 1148 1149
		if (r < 0)
			return r;
	}
A
Alasdair G Kergon 已提交
1150

1151 1152 1153
	bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
	clone->bi_iter.bi_size = to_bytes(len);

1154
	if (unlikely(bio_integrity(bio) != NULL))
1155
		bio_integrity_trim(clone, 0, len);
1156 1157

	return 0;
L
Linus Torvalds 已提交
1158 1159
}

1160
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1161
				      struct dm_target *ti,
1162
				      unsigned target_bio_nr)
1163
{
1164 1165 1166
	struct dm_target_io *tio;
	struct bio *clone;

1167
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1168
	tio = container_of(clone, struct dm_target_io, clone);
1169 1170 1171

	tio->io = ci->io;
	tio->ti = ti;
1172
	tio->target_bio_nr = target_bio_nr;
1173 1174 1175 1176

	return tio;
}

1177 1178
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1179
				       unsigned target_bio_nr, unsigned *len)
1180
{
1181
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1182
	struct bio *clone = &tio->clone;
1183

1184 1185
	tio->len_ptr = len;

1186
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1187
	if (len)
1188
		bio_setup_sector(clone, ci->sector, *len);
1189

A
Alasdair G Kergon 已提交
1190
	__map_bio(tio);
1191 1192
}

1193
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1194
				  unsigned num_bios, unsigned *len)
1195
{
1196
	unsigned target_bio_nr;
1197

1198
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1199
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1200 1201
}

1202
static int __send_empty_flush(struct clone_info *ci)
1203
{
1204
	unsigned target_nr = 0;
1205 1206
	struct dm_target *ti;

1207
	BUG_ON(bio_has_data(ci->bio));
1208
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1209
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1210 1211 1212 1213

	return 0;
}

1214
static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1215
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1216
{
1217
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1218
	struct dm_target_io *tio;
1219 1220
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
1221
	int r = 0;
M
Mike Snitzer 已提交
1222

1223 1224 1225 1226 1227
	/*
	 * 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
1228

1229
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1230
		tio = alloc_tio(ci, ti, target_bio_nr);
1231
		tio->len_ptr = len;
1232
		r = clone_bio(tio, bio, sector, *len);
1233
		if (r < 0) {
1234
			free_tio(tio);
1235
			break;
1236
		}
1237 1238
		__map_bio(tio);
	}
1239 1240

	return r;
M
Mike Snitzer 已提交
1241 1242
}

1243
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1244

1245
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1246
{
1247
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1248 1249
}

1250
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1251
{
1252
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1253 1254
}

1255 1256 1257 1258 1259
static unsigned get_num_write_zeroes_bios(struct dm_target *ti)
{
	return ti->num_write_zeroes_bios;
}

M
Mike Snitzer 已提交
1260
typedef bool (*is_split_required_fn)(struct dm_target *ti);
1261

M
Mike Snitzer 已提交
1262 1263
static bool is_split_required_for_discard(struct dm_target *ti)
{
1264
	return ti->split_discard_bios;
1265 1266
}

1267 1268 1269
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
1270
{
M
Mike Snitzer 已提交
1271
	struct dm_target *ti;
1272
	unsigned len;
1273
	unsigned num_bios;
1274

1275 1276 1277 1278
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
1279

M
Mike Snitzer 已提交
1280
		/*
M
Mike Snitzer 已提交
1281 1282
		 * 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
1283
		 * reconfiguration might also have changed that since the
1284
		 * check was performed.
M
Mike Snitzer 已提交
1285
		 */
1286 1287
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1288
			return -EOPNOTSUPP;
1289

M
Mike Snitzer 已提交
1290
		if (is_split_required && !is_split_required(ti))
1291
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1292
		else
1293
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1294

1295
		__send_duplicate_bios(ci, ti, num_bios, &len);
1296

1297 1298
		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1299 1300

	return 0;
1301 1302
}

1303
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1304
{
1305 1306
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1307
}
1308

1309
static int __send_write_same(struct clone_info *ci)
1310
{
1311
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
1312 1313
}

1314 1315 1316 1317 1318
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
1319 1320 1321
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1322
static int __split_and_process_non_flush(struct clone_info *ci)
1323
{
1324
	struct bio *bio = ci->bio;
1325
	struct dm_target *ti;
1326
	unsigned len;
1327
	int r;
1328

M
Mike Christie 已提交
1329
	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
1330
		return __send_discard(ci);
M
Mike Christie 已提交
1331
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME))
1332
		return __send_write_same(ci);
1333 1334
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES))
		return __send_write_zeroes(ci);
1335

1336 1337 1338 1339
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1340
	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
1341

1342 1343 1344
	r = __clone_and_map_data_bio(ci, ti, ci->sector, &len);
	if (r < 0)
		return r;
1345

1346 1347
	ci->sector += len;
	ci->sector_count -= len;
1348

1349
	return 0;
1350 1351
}

L
Linus Torvalds 已提交
1352
/*
1353
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1354
 */
M
Mikulas Patocka 已提交
1355 1356
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
1357
{
L
Linus Torvalds 已提交
1358
	struct clone_info ci;
1359
	int error = 0;
L
Linus Torvalds 已提交
1360

M
Mikulas Patocka 已提交
1361
	if (unlikely(!map)) {
1362
		bio_io_error(bio);
1363 1364
		return;
	}
1365

M
Mikulas Patocka 已提交
1366
	ci.map = map;
L
Linus Torvalds 已提交
1367 1368 1369 1370 1371 1372
	ci.md = md;
	ci.io = alloc_io(md);
	ci.io->error = 0;
	atomic_set(&ci.io->io_count, 1);
	ci.io->bio = bio;
	ci.io->md = md;
1373
	spin_lock_init(&ci.io->endio_lock);
1374
	ci.sector = bio->bi_iter.bi_sector;
1375

1376
	start_io_acct(ci.io);
1377

J
Jens Axboe 已提交
1378
	if (bio->bi_opf & REQ_PREFLUSH) {
1379 1380
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1381
		error = __send_empty_flush(&ci);
1382 1383
		/* dec_pending submits any data associated with flush */
	} else {
1384
		ci.bio = bio;
1385
		ci.sector_count = bio_sectors(bio);
1386
		while (ci.sector_count && !error)
1387
			error = __split_and_process_non_flush(&ci);
1388
	}
1389

L
Linus Torvalds 已提交
1390
	/* drop the extra reference count */
1391
	dec_pending(ci.io, error);
1392
}
L
Linus Torvalds 已提交
1393 1394 1395
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/
1396

1397
/*
L
Linus Torvalds 已提交
1398 1399
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
1400
 */
1401
static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio)
1402
{
1403
	int rw = bio_data_dir(bio);
1404
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1405 1406
	int srcu_idx;
	struct dm_table *map;
1407

M
Mikulas Patocka 已提交
1408
	map = dm_get_live_table(md, &srcu_idx);
1409

1410
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1411

1412 1413
	/* 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
1414
		dm_put_live_table(md, srcu_idx);
1415

J
Jens Axboe 已提交
1416
		if (!(bio->bi_opf & REQ_RAHEAD))
1417 1418
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1419
			bio_io_error(bio);
1420
		return BLK_QC_T_NONE;
1421
	}
L
Linus Torvalds 已提交
1422

M
Mikulas Patocka 已提交
1423 1424
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1425
	return BLK_QC_T_NONE;
1426 1427
}

L
Linus Torvalds 已提交
1428 1429
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1430 1431 1432
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1433

1434
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mike Snitzer 已提交
1435
		if (dm_request_based(md)) {
1436
			/*
M
Mike Snitzer 已提交
1437 1438
			 * With request-based DM we only need to check the
			 * top-level queue for congestion.
1439
			 */
1440
			r = md->queue->backing_dev_info->wb.state & bdi_bits;
M
Mike Snitzer 已提交
1441 1442 1443
		} else {
			map = dm_get_live_table_fast(md);
			if (map)
1444
				r = dm_table_any_congested(map, bdi_bits);
M
Mike Snitzer 已提交
1445
			dm_put_live_table_fast(md);
1446 1447 1448
		}
	}

L
Linus Torvalds 已提交
1449 1450 1451 1452 1453 1454
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1455
static void free_minor(int minor)
L
Linus Torvalds 已提交
1456
{
1457
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1458
	idr_remove(&_minor_idr, minor);
1459
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1460 1461 1462 1463 1464
}

/*
 * See if the device with a specific minor # is free.
 */
1465
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1466
{
T
Tejun Heo 已提交
1467
	int r;
L
Linus Torvalds 已提交
1468 1469 1470 1471

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

T
Tejun Heo 已提交
1472
	idr_preload(GFP_KERNEL);
1473
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1474

T
Tejun Heo 已提交
1475
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1476

1477
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1478 1479 1480 1481
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1482 1483
}

1484
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1485
{
T
Tejun Heo 已提交
1486
	int r;
J
Jeff Mahoney 已提交
1487

T
Tejun Heo 已提交
1488
	idr_preload(GFP_KERNEL);
1489
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1490

T
Tejun Heo 已提交
1491
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1492

1493
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1494 1495 1496 1497 1498
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1499 1500
}

1501
static const struct block_device_operations dm_blk_dops;
1502
static const struct dax_operations dm_dax_ops;
L
Linus Torvalds 已提交
1503

1504 1505
static void dm_wq_work(struct work_struct *work);

1506
void dm_init_md_queue(struct mapped_device *md)
1507 1508 1509
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
1510
	 * devices.  The type of this dm device may not have been decided yet.
1511 1512 1513 1514 1515 1516 1517
	 * 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);
1518 1519 1520 1521 1522 1523

	/*
	 * 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;
1524
	md->queue->backing_dev_info->congested_data = md;
1525
}
1526

1527
void dm_init_normal_md_queue(struct mapped_device *md)
1528
{
1529
	md->use_blk_mq = false;
1530 1531 1532 1533 1534
	dm_init_md_queue(md);

	/*
	 * Initialize aspects of queue that aren't relevant for blk-mq
	 */
1535
	md->queue->backing_dev_info->congested_fn = dm_any_congested;
1536 1537 1538
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

1539 1540 1541 1542 1543 1544
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);
1545
	mempool_destroy(md->io_pool);
1546 1547 1548
	if (md->bs)
		bioset_free(md->bs);

1549 1550 1551 1552 1553 1554
	if (md->dax_dev) {
		kill_dax(md->dax_dev);
		put_dax(md->dax_dev);
		md->dax_dev = NULL;
	}

1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
	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);

1566 1567
	cleanup_srcu_struct(&md->io_barrier);

1568 1569 1570 1571
	if (md->bdev) {
		bdput(md->bdev);
		md->bdev = NULL;
	}
1572 1573

	dm_mq_cleanup_mapped_device(md);
1574 1575
}

L
Linus Torvalds 已提交
1576 1577 1578
/*
 * Allocate and initialise a blank device with a given minor.
 */
1579
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1580
{
1581
	int r, numa_node_id = dm_get_numa_node();
1582
	struct dax_device *dax_dev;
1583
	struct mapped_device *md;
1584
	void *old_md;
L
Linus Torvalds 已提交
1585

1586
	md = kzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1587 1588 1589 1590 1591
	if (!md) {
		DMWARN("unable to allocate device, out of memory.");
		return NULL;
	}

1592
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1593
		goto bad_module_get;
1594

L
Linus Torvalds 已提交
1595
	/* get a minor number for the dev */
1596
	if (minor == DM_ANY_MINOR)
1597
		r = next_free_minor(&minor);
1598
	else
1599
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1600
	if (r < 0)
M
Milan Broz 已提交
1601
		goto bad_minor;
L
Linus Torvalds 已提交
1602

M
Mikulas Patocka 已提交
1603 1604 1605 1606
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

1607
	md->numa_node_id = numa_node_id;
1608
	md->use_blk_mq = dm_use_blk_mq_default();
1609
	md->init_tio_pdu = false;
1610
	md->type = DM_TYPE_NONE;
1611
	mutex_init(&md->suspend_lock);
1612
	mutex_init(&md->type_lock);
1613
	mutex_init(&md->table_devices_lock);
1614
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1615
	atomic_set(&md->holders, 1);
1616
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1617
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1618 1619
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
1620
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
1621
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1622

1623
	md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1624
	if (!md->queue)
1625
		goto bad;
L
Linus Torvalds 已提交
1626

1627
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1628

1629
	md->disk = alloc_disk_node(1, numa_node_id);
L
Linus Torvalds 已提交
1630
	if (!md->disk)
1631
		goto bad;
L
Linus Torvalds 已提交
1632

1633 1634
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1635
	init_waitqueue_head(&md->wait);
1636
	INIT_WORK(&md->work, dm_wq_work);
1637
	init_waitqueue_head(&md->eventq);
1638
	init_completion(&md->kobj_holder.completion);
1639
	md->kworker_task = NULL;
1640

L
Linus Torvalds 已提交
1641 1642 1643 1644 1645 1646
	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);
1647 1648 1649 1650 1651 1652

	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
1653
	add_disk(md->disk);
M
Mike Anderson 已提交
1654
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1655

T
Tejun Heo 已提交
1656
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
1657
	if (!md->wq)
1658
		goto bad;
1659

M
Mikulas Patocka 已提交
1660 1661
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
1662
		goto bad;
M
Mikulas Patocka 已提交
1663

1664
	bio_init(&md->flush_bio, NULL, 0);
1665
	md->flush_bio.bi_bdev = md->bdev;
1666
	md->flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
1667

M
Mikulas Patocka 已提交
1668 1669
	dm_stats_init(&md->stats);

1670
	/* Populate the mapping, nobody knows we exist yet */
1671
	spin_lock(&_minor_lock);
1672
	old_md = idr_replace(&_minor_idr, md, minor);
1673
	spin_unlock(&_minor_lock);
1674 1675 1676

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1677 1678
	return md;

1679 1680
bad:
	cleanup_mapped_device(md);
M
Mikulas Patocka 已提交
1681
bad_io_barrier:
L
Linus Torvalds 已提交
1682
	free_minor(minor);
M
Milan Broz 已提交
1683
bad_minor:
1684
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1685
bad_module_get:
L
Linus Torvalds 已提交
1686 1687 1688 1689
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1690 1691
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1692 1693
static void free_dev(struct mapped_device *md)
{
1694
	int minor = MINOR(disk_devt(md->disk));
1695

M
Mikulas Patocka 已提交
1696
	unlock_fs(md);
1697

1698
	cleanup_mapped_device(md);
1699

1700
	free_table_devices(&md->table_devices);
1701 1702 1703
	dm_stats_cleanup(&md->stats);
	free_minor(minor);

1704
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1705 1706 1707
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1708 1709
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1710
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1711

1712 1713
	if (md->bs) {
		/* The md already has necessary mempools. */
1714
		if (dm_table_bio_based(t)) {
1715 1716 1717 1718 1719 1720 1721 1722
			/*
			 * 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;
		}
1723 1724 1725 1726 1727 1728 1729 1730 1731
		/*
		 * 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
1732
	}
K
Kiyoshi Ueda 已提交
1733

1734
	BUG_ON(!p || md->io_pool || md->bs);
1735

K
Kiyoshi Ueda 已提交
1736 1737 1738 1739
	md->io_pool = p->io_pool;
	p->io_pool = NULL;
	md->bs = p->bs;
	p->bs = NULL;
1740

K
Kiyoshi Ueda 已提交
1741
out:
1742
	/* mempool bind completed, no longer need any mempools in the table */
K
Kiyoshi Ueda 已提交
1743 1744 1745
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
1746 1747 1748 1749 1750
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
1751 1752
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
1753 1754
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
1755 1756 1757 1758
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

1759
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
1760

L
Linus Torvalds 已提交
1761 1762 1763 1764
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

1765 1766 1767
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
1768
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
1769
{
1770 1771
	lockdep_assert_held(&md->suspend_lock);

1772
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
1773

1774
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
1775 1776
}

1777 1778 1779 1780 1781
/*
 * 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
1782
{
1783
	struct dm_table *old_map;
1784
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
1785 1786
	sector_t size;

1787 1788
	lockdep_assert_held(&md->suspend_lock);

L
Linus Torvalds 已提交
1789
	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
1790 1791 1792 1793

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
1794
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
1795 1796
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
1797
	__set_size(md, size);
1798

1799 1800
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
1801 1802 1803 1804 1805 1806 1807
	/*
	 * 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
1808
	if (dm_table_request_based(t)) {
1809
		dm_stop_queue(q);
M
Mike Snitzer 已提交
1810 1811 1812 1813 1814 1815 1816
		/*
		 * 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
1817 1818 1819

	__bind_mempools(md, t);

1820
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
1821
	rcu_assign_pointer(md->map, (void *)t);
1822 1823
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

1824
	dm_table_set_restrictions(t, q, limits);
1825 1826
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
1827

1828
	return old_map;
L
Linus Torvalds 已提交
1829 1830
}

1831 1832 1833 1834
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
1835
{
1836
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
1837 1838

	if (!map)
1839
		return NULL;
L
Linus Torvalds 已提交
1840 1841

	dm_table_event_callback(map, NULL, NULL);
1842
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
1843
	dm_sync_table(md);
1844 1845

	return map;
L
Linus Torvalds 已提交
1846 1847 1848 1849 1850
}

/*
 * Constructor for a new device.
 */
1851
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
1852 1853 1854
{
	struct mapped_device *md;

1855
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
1856 1857 1858
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
1859 1860
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
1861 1862 1863 1864
	*result = md;
	return 0;
}

1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878
/*
 * 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);
}

1879
void dm_set_md_type(struct mapped_device *md, enum dm_queue_mode type)
1880
{
1881
	BUG_ON(!mutex_is_locked(&md->type_lock));
1882 1883 1884
	md->type = type;
}

1885
enum dm_queue_mode dm_get_md_type(struct mapped_device *md)
1886 1887 1888 1889
{
	return md->type;
}

1890 1891 1892 1893 1894
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
/*
 * 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);

1906 1907 1908
/*
 * Setup the DM device's queue based on md's type
 */
1909
int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t)
1910
{
1911
	int r;
1912
	enum dm_queue_mode type = dm_get_md_type(md);
1913

1914
	switch (type) {
1915
	case DM_TYPE_REQUEST_BASED:
1916
		r = dm_old_init_request_queue(md, t);
1917
		if (r) {
1918
			DMERR("Cannot initialize queue for request-based mapped device");
1919
			return r;
1920
		}
1921 1922
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
1923
		r = dm_mq_init_request_queue(md, t);
1924
		if (r) {
1925
			DMERR("Cannot initialize queue for request-based dm-mq mapped device");
1926 1927 1928 1929
			return r;
		}
		break;
	case DM_TYPE_BIO_BASED:
1930
	case DM_TYPE_DAX_BIO_BASED:
1931
		dm_init_normal_md_queue(md);
1932
		blk_queue_make_request(md->queue, dm_make_request);
1933 1934 1935 1936 1937 1938
		/*
		 * 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;
1939 1940 1941

		if (type == DM_TYPE_DAX_BIO_BASED)
			queue_flag_set_unlocked(QUEUE_FLAG_DAX, md->queue);
1942
		break;
1943 1944 1945
	case DM_TYPE_NONE:
		WARN_ON_ONCE(true);
		break;
1946 1947 1948 1949 1950
	}

	return 0;
}

1951
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
1952 1953 1954 1955 1956 1957 1958
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

1959
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1960 1961

	md = idr_find(&_minor_idr, minor);
1962 1963 1964 1965 1966 1967 1968 1969 1970
	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
1971
	}
L
Linus Torvalds 已提交
1972

J
Jeff Mahoney 已提交
1973
out:
1974
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1975

1976 1977
	return md;
}
A
Alasdair G Kergon 已提交
1978
EXPORT_SYMBOL_GPL(dm_get_md);
1979

A
Alasdair G Kergon 已提交
1980
void *dm_get_mdptr(struct mapped_device *md)
1981
{
A
Alasdair G Kergon 已提交
1982
	return md->interface_ptr;
L
Linus Torvalds 已提交
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
}

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);
1993
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
1994 1995
}

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
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);

2009 2010 2011 2012 2013 2014
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2015
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2016
{
2017
	struct request_queue *q = dm_get_md_queue(md);
M
Mike Anderson 已提交
2018
	struct dm_table *map;
M
Mikulas Patocka 已提交
2019
	int srcu_idx;
L
Linus Torvalds 已提交
2020

2021
	might_sleep();
J
Jeff Mahoney 已提交
2022

2023
	spin_lock(&_minor_lock);
2024 2025 2026
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);
2027

2028
	blk_set_queue_dying(q);
2029

2030
	if (dm_request_based(md) && md->kworker_task)
P
Petr Mladek 已提交
2031
		kthread_flush_worker(&md->kworker);
2032

2033 2034 2035 2036 2037
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
2038
	map = dm_get_live_table(md, &srcu_idx);
2039 2040 2041
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2042
	}
M
Mikulas Patocka 已提交
2043 2044
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);
2045
	mutex_unlock(&md->suspend_lock);
M
Mikulas Patocka 已提交
2046

2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077
	/*
	 * 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
2078
}
E
Edward Goggin 已提交
2079
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2080

2081
static int dm_wait_for_completion(struct mapped_device *md, long task_state)
2082 2083
{
	int r = 0;
2084
	DEFINE_WAIT(wait);
2085 2086

	while (1) {
2087
		prepare_to_wait(&md->wait, &wait, task_state);
2088

2089
		if (!md_in_flight(md))
2090 2091
			break;

2092
		if (signal_pending_state(task_state, current)) {
2093 2094 2095 2096 2097 2098
			r = -EINTR;
			break;
		}

		io_schedule();
	}
2099
	finish_wait(&md->wait, &wait);
2100

2101 2102 2103
	return r;
}

L
Linus Torvalds 已提交
2104 2105 2106
/*
 * Process the deferred bios
 */
2107
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2108
{
2109 2110
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2111
	struct bio *c;
M
Mikulas Patocka 已提交
2112 2113
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2114

M
Mikulas Patocka 已提交
2115
	map = dm_get_live_table(md, &srcu_idx);
2116

2117
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2118 2119 2120 2121
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2122
		if (!c)
A
Alasdair G Kergon 已提交
2123
			break;
2124

K
Kiyoshi Ueda 已提交
2125 2126
		if (dm_request_based(md))
			generic_make_request(c);
2127
		else
M
Mikulas Patocka 已提交
2128
			__split_and_process_bio(md, map, c);
2129
	}
M
Milan Broz 已提交
2130

M
Mikulas Patocka 已提交
2131
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2132 2133
}

2134
static void dm_queue_flush(struct mapped_device *md)
2135
{
2136
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2137
	smp_mb__after_atomic();
2138
	queue_work(md->wq, &md->work);
2139 2140
}

L
Linus Torvalds 已提交
2141
/*
2142
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2143
 */
2144
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2145
{
2146
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2147
	struct queue_limits limits;
2148
	int r;
L
Linus Torvalds 已提交
2149

2150
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2151 2152

	/* device must be suspended */
2153
	if (!dm_suspended_md(md))
2154
		goto out;
L
Linus Torvalds 已提交
2155

2156 2157 2158 2159 2160 2161 2162
	/*
	 * 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
2163
		live_map = dm_get_live_table_fast(md);
2164 2165
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2166
		dm_put_live_table_fast(md);
2167 2168
	}

2169 2170 2171 2172 2173 2174
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2175
	}
2176

2177
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2178

2179
out:
2180
	mutex_unlock(&md->suspend_lock);
2181
	return map;
L
Linus Torvalds 已提交
2182 2183 2184 2185 2186 2187
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2188
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2189
{
2190
	int r;
L
Linus Torvalds 已提交
2191 2192

	WARN_ON(md->frozen_sb);
2193

2194
	md->frozen_sb = freeze_bdev(md->bdev);
2195
	if (IS_ERR(md->frozen_sb)) {
2196
		r = PTR_ERR(md->frozen_sb);
2197 2198
		md->frozen_sb = NULL;
		return r;
2199 2200
	}

2201 2202
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2203 2204 2205
	return 0;
}

2206
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2207
{
2208 2209 2210
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2211
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2212
	md->frozen_sb = NULL;
2213
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2214 2215 2216
}

/*
2217 2218 2219 2220
 * @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
 *
2221 2222 2223
 * 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.
2224
 */
2225
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
2226
			unsigned suspend_flags, long task_state,
2227
			int dmf_suspended_flag)
L
Linus Torvalds 已提交
2228
{
2229 2230 2231
	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
2232

2233 2234
	lockdep_assert_held(&md->suspend_lock);

2235 2236 2237 2238 2239 2240
	/*
	 * 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);
2241 2242
	else
		pr_debug("%s: suspending with flush\n", dm_device_name(md));
2243

2244 2245 2246 2247
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
2248 2249
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2250
	/*
K
Kiyoshi Ueda 已提交
2251 2252 2253 2254
	 * 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
2255 2256 2257
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
2258 2259
		if (r) {
			dm_table_presuspend_undo_targets(map);
2260
			return r;
2261
		}
2262
	}
L
Linus Torvalds 已提交
2263 2264

	/*
2265 2266 2267 2268 2269 2270 2271
	 * 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
2272 2273 2274
	 * __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
2275
	 */
2276
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2277 2278
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2279

2280
	/*
2281 2282
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2283
	 */
2284
	if (dm_request_based(md)) {
2285
		dm_stop_queue(md->queue);
2286
		if (md->kworker_task)
P
Petr Mladek 已提交
2287
			kthread_flush_worker(&md->kworker);
2288
	}
2289

2290 2291
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2292
	/*
2293 2294 2295
	 * 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
2296
	 */
2297
	r = dm_wait_for_completion(md, task_state);
2298 2299
	if (!r)
		set_bit(dmf_suspended_flag, &md->flags);
L
Linus Torvalds 已提交
2300

2301
	if (noflush)
2302
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2303 2304
	if (map)
		synchronize_srcu(&md->io_barrier);
2305

L
Linus Torvalds 已提交
2306
	/* were we interrupted ? */
2307
	if (r < 0) {
2308
		dm_queue_flush(md);
M
Milan Broz 已提交
2309

2310
		if (dm_request_based(md))
2311
			dm_start_queue(md->queue);
2312

2313
		unlock_fs(md);
2314
		dm_table_presuspend_undo_targets(map);
2315
		/* pushback list is already flushed, so skip flush */
2316
	}
L
Linus Torvalds 已提交
2317

2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
	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;
	}

2359
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2360

2361
	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED);
2362 2363
	if (r)
		goto out_unlock;
2364

2365 2366
	dm_table_postsuspend_targets(map);

2367
out_unlock:
2368
	mutex_unlock(&md->suspend_lock);
2369
	return r;
L
Linus Torvalds 已提交
2370 2371
}

2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
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))
2388
		dm_start_queue(md->queue);
2389 2390 2391 2392 2393 2394

	unlock_fs(md);

	return 0;
}

L
Linus Torvalds 已提交
2395 2396
int dm_resume(struct mapped_device *md)
{
2397
	int r;
2398
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2399

2400
retry:
2401
	r = -EINVAL;
2402 2403
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

2404
	if (!dm_suspended_md(md))
2405 2406
		goto out;

2407 2408 2409 2410 2411 2412 2413 2414 2415
	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;
	}

2416
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2417
	if (!map || !dm_table_get_size(map))
2418
		goto out;
L
Linus Torvalds 已提交
2419

2420
	r = __dm_resume(md, map);
2421 2422
	if (r)
		goto out;
2423 2424

	clear_bit(DMF_SUSPENDED, &md->flags);
2425
out:
2426
	mutex_unlock(&md->suspend_lock);
2427

2428
	return r;
L
Linus Torvalds 已提交
2429 2430
}

M
Mikulas Patocka 已提交
2431 2432 2433 2434 2435 2436
/*
 * 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.
 */

2437
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
2438
{
2439 2440
	struct dm_table *map = NULL;

2441 2442
	lockdep_assert_held(&md->suspend_lock);

2443
	if (md->internal_suspend_count++)
2444 2445 2446 2447 2448 2449 2450
		return; /* nested internal suspend */

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

2451
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2452 2453 2454 2455 2456 2457 2458

	/*
	 * 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.
	 */
2459 2460
	(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE,
			    DMF_SUSPENDED_INTERNALLY);
2461 2462 2463 2464 2465 2466

	dm_table_postsuspend_targets(map);
}

static void __dm_internal_resume(struct mapped_device *md)
{
2467 2468 2469
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
2470 2471
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
2472
	if (dm_suspended_md(md))
2473 2474 2475 2476 2477 2478 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
		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
2512 2513 2514 2515 2516 2517 2518
		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);
}
2519
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
2520

2521
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
2522
{
2523
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
2524 2525 2526 2527 2528 2529 2530
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
2531
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
2532

L
Linus Torvalds 已提交
2533 2534 2535
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2536
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2537
		       unsigned cookie)
2538
{
M
Milan Broz 已提交
2539 2540 2541 2542
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2543
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2544 2545 2546
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2547 2548
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2549
	}
2550 2551
}

M
Mike Anderson 已提交
2552 2553 2554 2555 2556
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567
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
2568 2569 2570 2571 2572 2573 2574 2575 2576
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
2577 2578 2579 2580 2581 2582 2583 2584
/*
 * 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;
}
2585
EXPORT_SYMBOL_GPL(dm_disk);
L
Linus Torvalds 已提交
2586

M
Milan Broz 已提交
2587 2588
struct kobject *dm_kobject(struct mapped_device *md)
{
2589
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
2590 2591 2592 2593 2594 2595
}

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

2596
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
2597

2598
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2599
	    dm_deleting_md(md))
2600 2601
		return NULL;

M
Milan Broz 已提交
2602 2603 2604 2605
	dm_get(md);
	return md;
}

2606
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2607 2608 2609 2610
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2611 2612 2613 2614 2615
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
2616 2617 2618 2619 2620
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

2621 2622
int dm_suspended(struct dm_target *ti)
{
2623
	return dm_suspended_md(dm_table_get_md(ti->table));
2624 2625 2626
}
EXPORT_SYMBOL_GPL(dm_suspended);

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

2633
struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type,
2634
					    unsigned integrity, unsigned per_io_data_size)
K
Kiyoshi Ueda 已提交
2635
{
2636
	struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id);
2637
	unsigned int pool_size = 0;
J
Jun'ichi Nomura 已提交
2638
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2639 2640

	if (!pools)
2641
		return NULL;
K
Kiyoshi Ueda 已提交
2642

2643 2644
	switch (type) {
	case DM_TYPE_BIO_BASED:
2645
	case DM_TYPE_DAX_BIO_BASED:
2646
		pool_size = dm_get_reserved_bio_based_ios();
2647
		front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
2648 2649 2650 2651
	
		pools->io_pool = mempool_create_slab_pool(pool_size, _io_cache);
		if (!pools->io_pool)
			goto out;
2652 2653 2654
		break;
	case DM_TYPE_REQUEST_BASED:
	case DM_TYPE_MQ_REQUEST_BASED:
2655
		pool_size = dm_get_reserved_rq_based_ios();
2656
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
2657
		/* per_io_data_size is used for blk-mq pdu at queue allocation */
2658 2659 2660 2661 2662
		break;
	default:
		BUG();
	}

J
Junichi Nomura 已提交
2663
	pools->bs = bioset_create_nobvec(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
2664
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2665
		goto out;
K
Kiyoshi Ueda 已提交
2666

2667
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2668
		goto out;
2669

K
Kiyoshi Ueda 已提交
2670
	return pools;
2671 2672 2673

out:
	dm_free_md_mempools(pools);
2674

2675
	return NULL;
K
Kiyoshi Ueda 已提交
2676 2677 2678 2679 2680 2681 2682
}

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

2683
	mempool_destroy(pools->io_pool);
2684

K
Kiyoshi Ueda 已提交
2685 2686 2687 2688 2689 2690
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

2691 2692 2693 2694 2695 2696 2697 2698 2699
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)
2700 2701
{
	struct mapped_device *md = bdev->bd_disk->private_data;
2702 2703 2704
	struct dm_table *table;
	struct dm_target *ti;
	int ret = -ENOTTY, srcu_idx;
2705

2706 2707 2708
	table = dm_get_live_table(md, &srcu_idx);
	if (!table || !dm_table_get_size(table))
		goto out;
2709

2710 2711 2712 2713
	/* 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);
2714

2715 2716 2717 2718 2719 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
	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;
2761 2762 2763
}

static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2764
			 u32 flags)
2765 2766 2767 2768
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2769
	int r;
2770

2771
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2772 2773 2774 2775 2776 2777 2778 2779 2780
	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;

2781
	bdput(bdev);
2782 2783 2784 2785 2786 2787 2788 2789
	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;
2790
	int r;
2791

2792
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2793 2794 2795 2796 2797 2798 2799 2800 2801
	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;

2802
	bdput(bdev);
2803 2804 2805 2806
	return r;
}

static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2807
			 enum pr_type type, bool abort)
2808 2809 2810 2811
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2812
	int r;
2813

2814
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2815 2816 2817 2818 2819 2820 2821 2822 2823
	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;

2824
	bdput(bdev);
2825 2826 2827 2828 2829 2830 2831 2832
	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;
2833
	int r;
2834

2835
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2836 2837 2838 2839 2840 2841 2842 2843 2844
	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;

2845
	bdput(bdev);
2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
	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,
};

2857
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2858 2859
	.open = dm_blk_open,
	.release = dm_blk_close,
2860
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2861
	.getgeo = dm_blk_getgeo,
2862
	.pr_ops = &dm_pr_ops,
L
Linus Torvalds 已提交
2863 2864 2865
	.owner = THIS_MODULE
};

2866 2867 2868 2869
static const struct dax_operations dm_dax_ops = {
	.direct_access = dm_dax_direct_access,
};

L
Linus Torvalds 已提交
2870 2871 2872 2873 2874 2875 2876 2877
/*
 * 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");
2878

2879 2880 2881
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

2882 2883 2884
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
2885 2886 2887
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");