dm.c 62.0 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>
L
Linus Torvalds 已提交
15 16 17 18 19
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/mempool.h>
#include <linux/slab.h>
#include <linux/idr.h>
D
Darrick J. Wong 已提交
20
#include <linux/hdreg.h>
21
#include <linux/delay.h>
22
#include <linux/wait.h>
23
#include <linux/pr.h>
24

25 26
#define DM_MSG_PREFIX "core"

N
Namhyung Kim 已提交
27 28 29 30 31 32 33 34 35 36
#ifdef CONFIG_PRINTK
/*
 * ratelimit state to be used in DMXXX_LIMIT().
 */
DEFINE_RATELIMIT_STATE(dm_ratelimit_state,
		       DEFAULT_RATELIMIT_INTERVAL,
		       DEFAULT_RATELIMIT_BURST);
EXPORT_SYMBOL(dm_ratelimit_state);
#endif

M
Milan Broz 已提交
37 38 39 40 41 42 43
/*
 * Cookies are numeric values sent with CHANGE and REMOVE
 * uevents while resuming, removing or renaming the device.
 */
#define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE"
#define DM_COOKIE_LENGTH 24

L
Linus Torvalds 已提交
44 45 46 47 48
static const char *_name = DM_NAME;

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

49 50
static DEFINE_IDR(_minor_idr);

51
static DEFINE_SPINLOCK(_minor_lock);
M
Mikulas Patocka 已提交
52 53 54 55 56

static void do_deferred_remove(struct work_struct *w);

static DECLARE_WORK(deferred_remove_work, do_deferred_remove);

57 58
static struct workqueue_struct *deferred_remove_workqueue;

L
Linus Torvalds 已提交
59 60 61 62 63 64 65
/*
 * 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
66
	struct bio *bio;
67
	unsigned long start_time;
68
	spinlock_t endio_lock;
M
Mikulas Patocka 已提交
69
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
70 71
};

72 73
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
74 75 76
/*
 * Bits for the md->flags field.
 */
77
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
78
#define DMF_SUSPENDED 1
79
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
80
#define DMF_FREEING 3
81
#define DMF_DELETING 4
82
#define DMF_NOFLUSH_SUSPENDING 5
83 84
#define DMF_DEFERRED_REMOVE 6
#define DMF_SUSPENDED_INTERNALLY 7
L
Linus Torvalds 已提交
85

86 87
#define DM_NUMA_NODE NUMA_NO_NODE
static int dm_numa_node = DM_NUMA_NODE;
88

K
Kiyoshi Ueda 已提交
89 90 91 92 93
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
94
	mempool_t *rq_pool;
K
Kiyoshi Ueda 已提交
95 96 97
	struct bio_set *bs;
};

98 99 100 101 102 103
struct table_device {
	struct list_head list;
	atomic_t count;
	struct dm_dev dm_dev;
};

104
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
105
static struct kmem_cache *_rq_tio_cache;
106
static struct kmem_cache *_rq_cache;
107

108 109 110
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
111
#define RESERVED_BIO_BASED_IOS		16
112 113
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
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;
}

135 136
unsigned __dm_get_module_param(unsigned *module_param,
			       unsigned def, unsigned max)
137
{
138 139
	unsigned param = ACCESS_ONCE(*module_param);
	unsigned modified_param = 0;
140

141 142 143 144
	if (!param)
		modified_param = def;
	else if (param > max)
		modified_param = max;
145

146 147 148
	if (modified_param) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
149 150
	}

151
	return param;
152 153
}

154 155
unsigned dm_get_reserved_bio_based_ios(void)
{
156
	return __dm_get_module_param(&reserved_bio_based_ios,
157
				     RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS);
158 159 160
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

161 162 163 164 165 166
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
167 168
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
169
	int r = -ENOMEM;
L
Linus Torvalds 已提交
170 171

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
172
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
173
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
174
		return r;
L
Linus Torvalds 已提交
175

K
Kiyoshi Ueda 已提交
176 177
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
178
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
179

180
	_rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request),
181 182 183 184
				      __alignof__(struct request), 0, NULL);
	if (!_rq_cache)
		goto out_free_rq_tio_cache;

M
Mike Anderson 已提交
185
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
186
	if (r)
187
		goto out_free_rq_cache;
M
Mike Anderson 已提交
188

189 190 191 192 193 194
	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
195 196
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
197
	if (r < 0)
198
		goto out_free_workqueue;
L
Linus Torvalds 已提交
199 200 201 202 203

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
204

205 206
out_free_workqueue:
	destroy_workqueue(deferred_remove_workqueue);
K
Kiyoshi Ueda 已提交
207 208
out_uevent_exit:
	dm_uevent_exit();
209 210
out_free_rq_cache:
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
211 212
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
213 214 215 216
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
217 218 219 220
}

static void local_exit(void)
{
M
Mikulas Patocka 已提交
221
	flush_scheduled_work();
222
	destroy_workqueue(deferred_remove_workqueue);
M
Mikulas Patocka 已提交
223

224
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
225
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
226
	kmem_cache_destroy(_io_cache);
227
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
228
	dm_uevent_exit();
L
Linus Torvalds 已提交
229 230 231 232 233 234

	_major = 0;

	DMINFO("cleaned up");
}

235
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
236 237 238 239
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
240
	dm_io_init,
241
	dm_kcopyd_init,
L
Linus Torvalds 已提交
242
	dm_interface_init,
M
Mikulas Patocka 已提交
243
	dm_statistics_init,
L
Linus Torvalds 已提交
244 245
};

246
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
247 248 249 250
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
251
	dm_io_exit,
252
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
253
	dm_interface_exit,
M
Mikulas Patocka 已提交
254
	dm_statistics_exit,
L
Linus Torvalds 已提交
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
};

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]();
284 285 286 287 288

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
289 290 291 292 293
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
294 295 296 297 298
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
299
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
300 301 302
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
303 304
	spin_lock(&_minor_lock);

A
Al Viro 已提交
305
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
306 307 308
	if (!md)
		goto out;

309
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
310
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
311 312 313 314
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
315
	dm_get(md);
316
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
317 318 319 320
out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
321 322
}

323
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
324
{
325
	struct mapped_device *md;
326

327 328
	spin_lock(&_minor_lock);

329 330 331 332
	md = disk->private_data;
	if (WARN_ON(!md))
		goto out;

M
Mikulas Patocka 已提交
333 334
	if (atomic_dec_and_test(&md->open_count) &&
	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
335
		queue_work(deferred_remove_workqueue, &deferred_remove_work);
M
Mikulas Patocka 已提交
336

L
Linus Torvalds 已提交
337
	dm_put(md);
338
out:
339
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
340 341
}

342 343 344 345 346 347 348 349
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
350
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
351 352 353 354 355
{
	int r = 0;

	spin_lock(&_minor_lock);

M
Mikulas Patocka 已提交
356
	if (dm_open_count(md)) {
357
		r = -EBUSY;
M
Mikulas Patocka 已提交
358 359 360 361
		if (mark_deferred)
			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
		r = -EEXIST;
362 363 364 365 366 367 368 369
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

M
Mikulas Patocka 已提交
370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
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
391 392 393 394 395
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

396 397 398 399 400
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
	return md->queue;
}

M
Mikulas Patocka 已提交
401 402 403 404 405
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
406 407 408 409 410 411 412
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);
}

413 414 415
static int dm_grab_bdev_for_ioctl(struct mapped_device *md,
				  struct block_device **bdev,
				  fmode_t *mode)
416
{
417
	struct dm_target *tgt;
418
	struct dm_table *map;
419
	int srcu_idx, r;
420

421
retry:
C
Christoph Hellwig 已提交
422
	r = -ENOTTY;
423
	map = dm_get_live_table(md, &srcu_idx);
424 425 426 427 428 429 430
	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;

431 432
	tgt = dm_table_get_target(map, 0);
	if (!tgt->type->prepare_ioctl)
433
		goto out;
434

435
	if (dm_suspended_md(md)) {
436 437 438 439
		r = -EAGAIN;
		goto out;
	}

440
	r = tgt->type->prepare_ioctl(tgt, bdev, mode);
C
Christoph Hellwig 已提交
441 442
	if (r < 0)
		goto out;
443

444 445
	bdgrab(*bdev);
	dm_put_live_table(md, srcu_idx);
C
Christoph Hellwig 已提交
446
	return r;
447 448

out:
449
	dm_put_live_table(md, srcu_idx);
450
	if (r == -ENOTCONN && !fatal_signal_pending(current)) {
451 452 453
		msleep(10);
		goto retry;
	}
C
Christoph Hellwig 已提交
454 455 456 457 458 459 460
	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;
461
	int r;
C
Christoph Hellwig 已提交
462

463
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
C
Christoph Hellwig 已提交
464 465
	if (r < 0)
		return r;
466

C
Christoph Hellwig 已提交
467 468 469 470 471 472 473 474 475 476
	if (r > 0) {
		/*
		 * Target determined this ioctl is being issued against
		 * a logical partition of the parent bdev; so extra
		 * validation is needed.
		 */
		r = scsi_verify_blk_ioctl(NULL, cmd);
		if (r)
			goto out;
	}
477

478
	r =  __blkdev_driver_ioctl(bdev, mode, cmd, arg);
C
Christoph Hellwig 已提交
479
out:
480
	bdput(bdev);
481 482 483
	return r;
}

A
Alasdair G Kergon 已提交
484
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
485 486 487 488
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
489
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
490 491 492 493
{
	mempool_free(io, md->io_pool);
}

494
static void free_tio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
495
{
496
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
497 498
}

499
int md_in_flight(struct mapped_device *md)
K
Kiyoshi Ueda 已提交
500 501 502 503 504
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

505 506 507
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
508
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
509
	int cpu;
M
Mikulas Patocka 已提交
510
	int rw = bio_data_dir(bio);
511 512 513

	io->start_time = jiffies;

T
Tejun Heo 已提交
514 515 516
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
517 518
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
519 520

	if (unlikely(dm_stats_used(&md->stats)))
521 522 523
		dm_stats_account_io(&md->stats, bio_data_dir(bio),
				    bio->bi_iter.bi_sector, bio_sectors(bio),
				    false, 0, &io->stats_aux);
524 525
}

526
static void end_io_acct(struct dm_io *io)
527 528 529 530
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
531
	int pending;
532 533
	int rw = bio_data_dir(bio);

534
	generic_end_io_acct(rw, &dm_disk(md)->part0, io->start_time);
535

M
Mikulas Patocka 已提交
536
	if (unlikely(dm_stats_used(&md->stats)))
537 538 539
		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
540

541 542
	/*
	 * After this is decremented the bio must not be touched if it is
543
	 * a flush.
544
	 */
545 546
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
547
	pending += atomic_read(&md->pending[rw^0x1]);
548

549 550 551
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
552 553
}

L
Linus Torvalds 已提交
554 555 556
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
557
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
558
{
559
	unsigned long flags;
L
Linus Torvalds 已提交
560

561
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
562
	bio_list_add(&md->deferred, bio);
563
	spin_unlock_irqrestore(&md->deferred_lock, flags);
564
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
565 566 567 568 569
}

/*
 * 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
570
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
571
 */
M
Mikulas Patocka 已提交
572
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
573
{
M
Mikulas Patocka 已提交
574 575 576 577
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
578

M
Mikulas Patocka 已提交
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
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
599

M
Mikulas Patocka 已提交
600 601 602
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
603 604
}

605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663
/*
 * 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;
	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);
	td->dm_dev.bdev = NULL;
}

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) {
664
		td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id);
665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
		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
719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
/*
 * 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
746 747 748 749 750 751 752 753 754
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

755 756 757 758 759
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
760 761 762 763
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
764
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
765
{
766
	unsigned long flags;
767 768 769
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
770 771

	/* Push-back supersedes any I/O errors */
772 773 774 775 776 777
	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
778 779

	if (atomic_dec_and_test(&io->io_count)) {
780 781 782 783
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
784
			spin_lock_irqsave(&md->deferred_lock, flags);
785 786 787
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
788 789
				/* noflush suspend was interrupted. */
				io->error = -EIO;
790
			spin_unlock_irqrestore(&md->deferred_lock, flags);
791 792
		}

793 794
		io_error = io->error;
		bio = io->bio;
795 796 797 798 799
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
800

801
		if ((bio->bi_rw & REQ_PREFLUSH) && bio->bi_iter.bi_size) {
802
			/*
803
			 * Preflush done for flush with data, reissue
804
			 * without REQ_PREFLUSH.
805
			 */
806
			bio->bi_rw &= ~REQ_PREFLUSH;
807
			queue_io(md, bio);
808
		} else {
809
			/* done with normal IO or empty flush */
810
			trace_block_bio_complete(md->queue, bio, io_error);
811 812
			bio->bi_error = io_error;
			bio_endio(bio);
813
		}
L
Linus Torvalds 已提交
814 815 816
	}
}

817
void disable_write_same(struct mapped_device *md)
818 819 820 821 822 823 824
{
	struct queue_limits *limits = dm_get_queue_limits(md);

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

825
static void clone_endio(struct bio *bio)
L
Linus Torvalds 已提交
826
{
827
	int error = bio->bi_error;
828
	int r = error;
M
Mikulas Patocka 已提交
829
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
830
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
831
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
832 833 834
	dm_endio_fn endio = tio->ti->type->end_io;

	if (endio) {
M
Mikulas Patocka 已提交
835
		r = endio(tio->ti, bio, error);
836 837 838 839 840
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
841
			error = r;
842 843
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
844
			return;
845 846 847 848
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
849 850
	}

M
Mike Christie 已提交
851
	if (unlikely(r == -EREMOTEIO && (bio_op(bio) == REQ_OP_WRITE_SAME) &&
852 853 854
		     !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors))
		disable_write_same(md);

855
	free_tio(tio);
856
	dec_pending(io, error);
L
Linus Torvalds 已提交
857 858
}

859 860 861 862 863 864 865 866 867 868 869 870
/*
 * 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
871
{
872
	sector_t len = max_io_len_target_boundary(sector, ti);
873
	sector_t offset, max_len;
L
Linus Torvalds 已提交
874 875

	/*
876
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
877
	 */
878 879 880 881 882 883 884 885 886 887
	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
888 889 890 891 892
	}

	return len;
}

893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
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);

908
static long dm_blk_direct_access(struct block_device *bdev, sector_t sector,
909
				 void **kaddr, pfn_t *pfn, long size)
910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	struct dm_table *map;
	struct dm_target *ti;
	int srcu_idx;
	long len, ret = -EIO;

	map = dm_get_live_table(md, &srcu_idx);
	if (!map)
		goto out;

	ti = dm_table_find_target(map, sector);
	if (!dm_target_is_valid(ti))
		goto out;

	len = max_io_len(sector, ti) << SECTOR_SHIFT;
	size = min(len, size);

	if (ti->type->direct_access)
		ret = ti->type->direct_access(ti, sector, kaddr, pfn, size);
out:
	dm_put_live_table(md, srcu_idx);
	return min(ret, size);
}

935 936
/*
 * A target may call dm_accept_partial_bio only from the map routine.  It is
937
 * allowed for all bio types except REQ_PREFLUSH.
938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
 *
 * 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;
967
	BUG_ON(bio->bi_rw & REQ_PREFLUSH);
968 969 970 971 972 973 974
	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);

A
Alasdair G Kergon 已提交
975
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
976 977
{
	int r;
978
	sector_t sector;
979
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
980
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
981 982 983 984 985 986 987 988 989

	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);
990
	sector = clone->bi_iter.bi_sector;
M
Mikulas Patocka 已提交
991
	r = ti->type->map(ti, clone);
992
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
993
		/* the bio has been remapped so dispatch it */
994

995 996
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
997

L
Linus Torvalds 已提交
998
		generic_make_request(clone);
999 1000
	} else if (r < 0 || r == DM_MAPIO_REQUEUE) {
		/* error the io and bail out, or requeue it if needed */
S
Stefan Bader 已提交
1001
		dec_pending(tio->io, r);
1002
		free_tio(tio);
1003
	} else if (r != DM_MAPIO_SUBMITTED) {
1004 1005
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1006 1007 1008 1009 1010 1011 1012 1013 1014
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1015
	unsigned sector_count;
L
Linus Torvalds 已提交
1016 1017
};

1018
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1019
{
1020 1021
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1027 1028
static int clone_bio(struct dm_target_io *tio, struct bio *bio,
		     sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1029
{
1030
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1031

1032 1033
	__bio_clone_fast(clone, bio);

1034 1035 1036 1037 1038
	if (bio_integrity(bio)) {
		int r = bio_integrity_clone(clone, bio, GFP_NOIO);
		if (r < 0)
			return r;
	}
A
Alasdair G Kergon 已提交
1039

1040 1041 1042 1043 1044
	bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
	clone->bi_iter.bi_size = to_bytes(len);

	if (bio_integrity(bio))
		bio_integrity_trim(clone, 0, len);
1045 1046

	return 0;
L
Linus Torvalds 已提交
1047 1048
}

1049
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1050
				      struct dm_target *ti,
1051
				      unsigned target_bio_nr)
1052
{
1053 1054 1055
	struct dm_target_io *tio;
	struct bio *clone;

1056
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1057
	tio = container_of(clone, struct dm_target_io, clone);
1058 1059 1060

	tio->io = ci->io;
	tio->ti = ti;
1061
	tio->target_bio_nr = target_bio_nr;
1062 1063 1064 1065

	return tio;
}

1066 1067
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1068
				       unsigned target_bio_nr, unsigned *len)
1069
{
1070
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1071
	struct bio *clone = &tio->clone;
1072

1073 1074
	tio->len_ptr = len;

1075
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1076
	if (len)
1077
		bio_setup_sector(clone, ci->sector, *len);
1078

A
Alasdair G Kergon 已提交
1079
	__map_bio(tio);
1080 1081
}

1082
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1083
				  unsigned num_bios, unsigned *len)
1084
{
1085
	unsigned target_bio_nr;
1086

1087
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1088
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1089 1090
}

1091
static int __send_empty_flush(struct clone_info *ci)
1092
{
1093
	unsigned target_nr = 0;
1094 1095
	struct dm_target *ti;

1096
	BUG_ON(bio_has_data(ci->bio));
1097
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1098
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1099 1100 1101 1102

	return 0;
}

1103
static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1104
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1105
{
1106
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1107
	struct dm_target_io *tio;
1108 1109
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
1110
	int r = 0;
M
Mike Snitzer 已提交
1111

1112 1113 1114 1115 1116
	/*
	 * 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
1117

1118
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1119
		tio = alloc_tio(ci, ti, target_bio_nr);
1120
		tio->len_ptr = len;
1121
		r = clone_bio(tio, bio, sector, *len);
1122
		if (r < 0) {
1123
			free_tio(tio);
1124
			break;
1125
		}
1126 1127
		__map_bio(tio);
	}
1128 1129

	return r;
M
Mike Snitzer 已提交
1130 1131
}

1132
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1133

1134
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1135
{
1136
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1137 1138
}

1139
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1140
{
1141
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1142 1143 1144
}

typedef bool (*is_split_required_fn)(struct dm_target *ti);
1145

M
Mike Snitzer 已提交
1146 1147
static bool is_split_required_for_discard(struct dm_target *ti)
{
1148
	return ti->split_discard_bios;
1149 1150
}

1151 1152 1153
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
1154
{
M
Mike Snitzer 已提交
1155
	struct dm_target *ti;
1156
	unsigned len;
1157
	unsigned num_bios;
1158

1159 1160 1161 1162
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
1163

M
Mike Snitzer 已提交
1164
		/*
M
Mike Snitzer 已提交
1165 1166
		 * 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
1167
		 * reconfiguration might also have changed that since the
1168
		 * check was performed.
M
Mike Snitzer 已提交
1169
		 */
1170 1171
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1172
			return -EOPNOTSUPP;
1173

M
Mike Snitzer 已提交
1174
		if (is_split_required && !is_split_required(ti))
1175
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1176
		else
1177
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1178

1179
		__send_duplicate_bios(ci, ti, num_bios, &len);
1180

1181 1182
		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1183 1184

	return 0;
1185 1186
}

1187
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1188
{
1189 1190
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1191
}
1192

1193
static int __send_write_same(struct clone_info *ci)
1194
{
1195
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
1196 1197
}

A
Alasdair G Kergon 已提交
1198 1199 1200
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1201
static int __split_and_process_non_flush(struct clone_info *ci)
1202
{
1203
	struct bio *bio = ci->bio;
1204
	struct dm_target *ti;
1205
	unsigned len;
1206
	int r;
1207

M
Mike Christie 已提交
1208
	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
1209
		return __send_discard(ci);
M
Mike Christie 已提交
1210
	else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME))
1211
		return __send_write_same(ci);
1212

1213 1214 1215 1216
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1217
	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
1218

1219 1220 1221
	r = __clone_and_map_data_bio(ci, ti, ci->sector, &len);
	if (r < 0)
		return r;
1222

1223 1224
	ci->sector += len;
	ci->sector_count -= len;
1225

1226
	return 0;
1227 1228
}

L
Linus Torvalds 已提交
1229
/*
1230
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1231
 */
M
Mikulas Patocka 已提交
1232 1233
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
1234
{
L
Linus Torvalds 已提交
1235
	struct clone_info ci;
1236
	int error = 0;
L
Linus Torvalds 已提交
1237

M
Mikulas Patocka 已提交
1238
	if (unlikely(!map)) {
1239
		bio_io_error(bio);
1240 1241
		return;
	}
1242

M
Mikulas Patocka 已提交
1243
	ci.map = map;
L
Linus Torvalds 已提交
1244 1245 1246 1247 1248 1249
	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;
1250
	spin_lock_init(&ci.io->endio_lock);
1251
	ci.sector = bio->bi_iter.bi_sector;
1252

1253
	start_io_acct(ci.io);
1254

1255
	if (bio->bi_rw & REQ_PREFLUSH) {
1256 1257
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1258
		error = __send_empty_flush(&ci);
1259 1260
		/* dec_pending submits any data associated with flush */
	} else {
1261
		ci.bio = bio;
1262
		ci.sector_count = bio_sectors(bio);
1263
		while (ci.sector_count && !error)
1264
			error = __split_and_process_non_flush(&ci);
1265
	}
1266

L
Linus Torvalds 已提交
1267
	/* drop the extra reference count */
1268
	dec_pending(ci.io, error);
1269
}
L
Linus Torvalds 已提交
1270 1271 1272
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/
1273

1274
/*
L
Linus Torvalds 已提交
1275 1276
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
1277
 */
1278
static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio)
1279
{
1280
	int rw = bio_data_dir(bio);
1281
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1282 1283
	int srcu_idx;
	struct dm_table *map;
1284

M
Mikulas Patocka 已提交
1285
	map = dm_get_live_table(md, &srcu_idx);
1286

1287
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1288

1289 1290
	/* 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
1291
		dm_put_live_table(md, srcu_idx);
1292

1293
		if (!(bio->bi_rw & REQ_RAHEAD))
1294 1295
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1296
			bio_io_error(bio);
1297
		return BLK_QC_T_NONE;
1298
	}
L
Linus Torvalds 已提交
1299

M
Mikulas Patocka 已提交
1300 1301
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1302
	return BLK_QC_T_NONE;
1303 1304
}

L
Linus Torvalds 已提交
1305 1306
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1307 1308 1309
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1310

1311
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mike Snitzer 已提交
1312
		if (dm_request_based(md)) {
1313
			/*
M
Mike Snitzer 已提交
1314 1315
			 * With request-based DM we only need to check the
			 * top-level queue for congestion.
1316
			 */
M
Mike Snitzer 已提交
1317 1318 1319 1320
			r = md->queue->backing_dev_info.wb.state & bdi_bits;
		} else {
			map = dm_get_live_table_fast(md);
			if (map)
1321
				r = dm_table_any_congested(map, bdi_bits);
M
Mike Snitzer 已提交
1322
			dm_put_live_table_fast(md);
1323 1324 1325
		}
	}

L
Linus Torvalds 已提交
1326 1327 1328 1329 1330 1331
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1332
static void free_minor(int minor)
L
Linus Torvalds 已提交
1333
{
1334
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1335
	idr_remove(&_minor_idr, minor);
1336
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1337 1338 1339 1340 1341
}

/*
 * See if the device with a specific minor # is free.
 */
1342
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1343
{
T
Tejun Heo 已提交
1344
	int r;
L
Linus Torvalds 已提交
1345 1346 1347 1348

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

T
Tejun Heo 已提交
1349
	idr_preload(GFP_KERNEL);
1350
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1351

T
Tejun Heo 已提交
1352
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1353

1354
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1355 1356 1357 1358
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1359 1360
}

1361
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1362
{
T
Tejun Heo 已提交
1363
	int r;
J
Jeff Mahoney 已提交
1364

T
Tejun Heo 已提交
1365
	idr_preload(GFP_KERNEL);
1366
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1367

T
Tejun Heo 已提交
1368
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1369

1370
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1371 1372 1373 1374 1375
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1376 1377
}

1378
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1379

1380 1381
static void dm_wq_work(struct work_struct *work);

1382
void dm_init_md_queue(struct mapped_device *md)
1383 1384 1385
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
1386
	 * devices.  The type of this dm device may not have been decided yet.
1387 1388 1389 1390 1391 1392 1393
	 * 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);
1394 1395 1396 1397 1398 1399 1400

	/*
	 * 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;
	md->queue->backing_dev_info.congested_data = md;
1401
}
1402

1403
void dm_init_normal_md_queue(struct mapped_device *md)
1404
{
1405
	md->use_blk_mq = false;
1406 1407 1408 1409 1410
	dm_init_md_queue(md);

	/*
	 * Initialize aspects of queue that aren't relevant for blk-mq
	 */
1411 1412 1413 1414
	md->queue->backing_dev_info.congested_fn = dm_any_congested;
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

1415 1416 1417 1418 1419 1420
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);
1421 1422
	mempool_destroy(md->io_pool);
	mempool_destroy(md->rq_pool);
1423 1424 1425
	if (md->bs)
		bioset_free(md->bs);

1426 1427
	cleanup_srcu_struct(&md->io_barrier);

1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
	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);

	if (md->bdev) {
		bdput(md->bdev);
		md->bdev = NULL;
	}
1443 1444

	dm_mq_cleanup_mapped_device(md);
1445 1446
}

L
Linus Torvalds 已提交
1447 1448 1449
/*
 * Allocate and initialise a blank device with a given minor.
 */
1450
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1451
{
1452 1453
	int r, numa_node_id = dm_get_numa_node();
	struct mapped_device *md;
1454
	void *old_md;
L
Linus Torvalds 已提交
1455

1456
	md = kzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1457 1458 1459 1460 1461
	if (!md) {
		DMWARN("unable to allocate device, out of memory.");
		return NULL;
	}

1462
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1463
		goto bad_module_get;
1464

L
Linus Torvalds 已提交
1465
	/* get a minor number for the dev */
1466
	if (minor == DM_ANY_MINOR)
1467
		r = next_free_minor(&minor);
1468
	else
1469
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1470
	if (r < 0)
M
Milan Broz 已提交
1471
		goto bad_minor;
L
Linus Torvalds 已提交
1472

M
Mikulas Patocka 已提交
1473 1474 1475 1476
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

1477
	md->numa_node_id = numa_node_id;
1478
	md->use_blk_mq = dm_use_blk_mq_default();
1479
	md->init_tio_pdu = false;
1480
	md->type = DM_TYPE_NONE;
1481
	mutex_init(&md->suspend_lock);
1482
	mutex_init(&md->type_lock);
1483
	mutex_init(&md->table_devices_lock);
1484
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1485
	atomic_set(&md->holders, 1);
1486
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1487
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1488 1489
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
1490
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
1491
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1492

1493
	md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id);
L
Linus Torvalds 已提交
1494
	if (!md->queue)
1495
		goto bad;
L
Linus Torvalds 已提交
1496

1497
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1498

1499
	md->disk = alloc_disk_node(1, numa_node_id);
L
Linus Torvalds 已提交
1500
	if (!md->disk)
1501
		goto bad;
L
Linus Torvalds 已提交
1502

1503 1504
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1505
	init_waitqueue_head(&md->wait);
1506
	INIT_WORK(&md->work, dm_wq_work);
1507
	init_waitqueue_head(&md->eventq);
1508
	init_completion(&md->kobj_holder.completion);
1509
	md->kworker_task = NULL;
1510

L
Linus Torvalds 已提交
1511 1512 1513 1514 1515 1516 1517
	md->disk->major = _major;
	md->disk->first_minor = minor;
	md->disk->fops = &dm_blk_dops;
	md->disk->queue = md->queue;
	md->disk->private_data = md;
	sprintf(md->disk->disk_name, "dm-%d", minor);
	add_disk(md->disk);
M
Mike Anderson 已提交
1518
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1519

T
Tejun Heo 已提交
1520
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
1521
	if (!md->wq)
1522
		goto bad;
1523

M
Mikulas Patocka 已提交
1524 1525
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
1526
		goto bad;
M
Mikulas Patocka 已提交
1527

1528 1529
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
M
Mike Christie 已提交
1530
	bio_set_op_attrs(&md->flush_bio, REQ_OP_WRITE, WRITE_FLUSH);
1531

M
Mikulas Patocka 已提交
1532 1533
	dm_stats_init(&md->stats);

1534
	/* Populate the mapping, nobody knows we exist yet */
1535
	spin_lock(&_minor_lock);
1536
	old_md = idr_replace(&_minor_idr, md, minor);
1537
	spin_unlock(&_minor_lock);
1538 1539 1540

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1541 1542
	return md;

1543 1544
bad:
	cleanup_mapped_device(md);
M
Mikulas Patocka 已提交
1545
bad_io_barrier:
L
Linus Torvalds 已提交
1546
	free_minor(minor);
M
Milan Broz 已提交
1547
bad_minor:
1548
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1549
bad_module_get:
L
Linus Torvalds 已提交
1550 1551 1552 1553
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1554 1555
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1556 1557
static void free_dev(struct mapped_device *md)
{
1558
	int minor = MINOR(disk_devt(md->disk));
1559

M
Mikulas Patocka 已提交
1560
	unlock_fs(md);
1561

1562
	cleanup_mapped_device(md);
1563

1564
	free_table_devices(&md->table_devices);
1565 1566 1567
	dm_stats_cleanup(&md->stats);
	free_minor(minor);

1568
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1569 1570 1571
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1572 1573
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1574
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1575

1576 1577
	if (md->bs) {
		/* The md already has necessary mempools. */
1578
		if (dm_table_bio_based(t)) {
1579 1580 1581 1582 1583 1584 1585 1586
			/*
			 * 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;
		}
1587 1588 1589 1590 1591 1592 1593 1594 1595
		/*
		 * 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
1596
	}
K
Kiyoshi Ueda 已提交
1597

1598 1599
	BUG_ON(!p || md->io_pool || md->rq_pool || md->bs);

K
Kiyoshi Ueda 已提交
1600 1601
	md->io_pool = p->io_pool;
	p->io_pool = NULL;
1602 1603
	md->rq_pool = p->rq_pool;
	p->rq_pool = NULL;
K
Kiyoshi Ueda 已提交
1604 1605
	md->bs = p->bs;
	p->bs = NULL;
1606

K
Kiyoshi Ueda 已提交
1607
out:
1608
	/* mempool bind completed, no longer need any mempools in the table */
K
Kiyoshi Ueda 已提交
1609 1610 1611
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
1612 1613 1614 1615 1616
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
1617 1618
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
1619 1620
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
1621 1622 1623 1624
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

1625
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
1626

L
Linus Torvalds 已提交
1627 1628 1629 1630
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

1631 1632 1633
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
1634
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
1635
{
1636
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
1637

1638
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
1639 1640
}

1641 1642 1643 1644 1645
/*
 * 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
1646
{
1647
	struct dm_table *old_map;
1648
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
1649 1650 1651
	sector_t size;

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
1652 1653 1654 1655

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
1656
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
1657 1658
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
1659
	__set_size(md, size);
1660

1661 1662
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
1663 1664 1665 1666 1667 1668 1669
	/*
	 * 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
1670
	if (dm_table_request_based(t)) {
1671
		dm_stop_queue(q);
M
Mike Snitzer 已提交
1672 1673 1674 1675 1676 1677 1678
		/*
		 * 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
1679 1680 1681

	__bind_mempools(md, t);

1682
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
1683
	rcu_assign_pointer(md->map, (void *)t);
1684 1685
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

1686
	dm_table_set_restrictions(t, q, limits);
1687 1688
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
1689

1690
	return old_map;
L
Linus Torvalds 已提交
1691 1692
}

1693 1694 1695 1696
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
1697
{
1698
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
1699 1700

	if (!map)
1701
		return NULL;
L
Linus Torvalds 已提交
1702 1703

	dm_table_event_callback(map, NULL, NULL);
1704
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
1705
	dm_sync_table(md);
1706 1707

	return map;
L
Linus Torvalds 已提交
1708 1709 1710 1711 1712
}

/*
 * Constructor for a new device.
 */
1713
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
1714 1715 1716
{
	struct mapped_device *md;

1717
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
1718 1719 1720
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
1721 1722
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
1723 1724 1725 1726
	*result = md;
	return 0;
}

1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
/*
 * Functions to manage md->type.
 * All are required to hold md->type_lock.
 */
void dm_lock_md_type(struct mapped_device *md)
{
	mutex_lock(&md->type_lock);
}

void dm_unlock_md_type(struct mapped_device *md)
{
	mutex_unlock(&md->type_lock);
}

void dm_set_md_type(struct mapped_device *md, unsigned type)
{
1743
	BUG_ON(!mutex_is_locked(&md->type_lock));
1744 1745 1746 1747 1748 1749 1750 1751
	md->type = type;
}

unsigned dm_get_md_type(struct mapped_device *md)
{
	return md->type;
}

1752 1753 1754 1755 1756
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767
/*
 * 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);

1768 1769 1770
/*
 * Setup the DM device's queue based on md's type
 */
1771
int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t)
1772
{
1773
	int r;
1774
	unsigned type = dm_get_md_type(md);
1775

1776
	switch (type) {
1777
	case DM_TYPE_REQUEST_BASED:
1778
		r = dm_old_init_request_queue(md);
1779
		if (r) {
1780
			DMERR("Cannot initialize queue for request-based mapped device");
1781
			return r;
1782
		}
1783 1784
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
1785
		r = dm_mq_init_request_queue(md, t);
1786
		if (r) {
1787
			DMERR("Cannot initialize queue for request-based dm-mq mapped device");
1788 1789 1790 1791
			return r;
		}
		break;
	case DM_TYPE_BIO_BASED:
1792
	case DM_TYPE_DAX_BIO_BASED:
1793
		dm_init_normal_md_queue(md);
1794
		blk_queue_make_request(md->queue, dm_make_request);
1795 1796 1797 1798 1799 1800
		/*
		 * 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;
1801 1802 1803

		if (type == DM_TYPE_DAX_BIO_BASED)
			queue_flag_set_unlocked(QUEUE_FLAG_DAX, md->queue);
1804
		break;
1805 1806 1807 1808 1809
	}

	return 0;
}

1810
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
1811 1812 1813 1814 1815 1816 1817
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

1818
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1819 1820

	md = idr_find(&_minor_idr, minor);
1821 1822 1823 1824 1825 1826 1827 1828 1829
	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
1830
	}
L
Linus Torvalds 已提交
1831

J
Jeff Mahoney 已提交
1832
out:
1833
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1834

1835 1836
	return md;
}
A
Alasdair G Kergon 已提交
1837
EXPORT_SYMBOL_GPL(dm_get_md);
1838

A
Alasdair G Kergon 已提交
1839
void *dm_get_mdptr(struct mapped_device *md)
1840
{
A
Alasdair G Kergon 已提交
1841
	return md->interface_ptr;
L
Linus Torvalds 已提交
1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
}

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);
1852
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
1853 1854
}

1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
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);

1868 1869 1870 1871 1872 1873
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

1874
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
1875
{
M
Mike Anderson 已提交
1876
	struct dm_table *map;
M
Mikulas Patocka 已提交
1877
	int srcu_idx;
L
Linus Torvalds 已提交
1878

1879
	might_sleep();
J
Jeff Mahoney 已提交
1880

1881
	spin_lock(&_minor_lock);
1882 1883 1884 1885
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);

1886
	if (dm_request_based(md) && md->kworker_task)
1887 1888
		flush_kthread_worker(&md->kworker);

1889 1890 1891 1892 1893
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
1894
	map = dm_get_live_table(md, &srcu_idx);
1895 1896 1897
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
1898
	}
M
Mikulas Patocka 已提交
1899 1900
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);
1901
	mutex_unlock(&md->suspend_lock);
M
Mikulas Patocka 已提交
1902

1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
	/*
	 * 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
1934
}
E
Edward Goggin 已提交
1935
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
1936

1937
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
1938 1939
{
	int r = 0;
1940 1941 1942
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
1943 1944

	while (1) {
1945
		set_current_state(interruptible);
1946

1947
		if (!md_in_flight(md))
1948 1949
			break;

1950 1951
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
1952 1953 1954 1955 1956 1957 1958 1959
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

1960 1961
	remove_wait_queue(&md->wait, &wait);

1962 1963 1964
	return r;
}

L
Linus Torvalds 已提交
1965 1966 1967
/*
 * Process the deferred bios
 */
1968
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
1969
{
1970 1971
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
1972
	struct bio *c;
M
Mikulas Patocka 已提交
1973 1974
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
1975

M
Mikulas Patocka 已提交
1976
	map = dm_get_live_table(md, &srcu_idx);
1977

1978
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
1979 1980 1981 1982
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

1983
		if (!c)
A
Alasdair G Kergon 已提交
1984
			break;
1985

K
Kiyoshi Ueda 已提交
1986 1987
		if (dm_request_based(md))
			generic_make_request(c);
1988
		else
M
Mikulas Patocka 已提交
1989
			__split_and_process_bio(md, map, c);
1990
	}
M
Milan Broz 已提交
1991

M
Mikulas Patocka 已提交
1992
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
1993 1994
}

1995
static void dm_queue_flush(struct mapped_device *md)
1996
{
1997
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
1998
	smp_mb__after_atomic();
1999
	queue_work(md->wq, &md->work);
2000 2001
}

L
Linus Torvalds 已提交
2002
/*
2003
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2004
 */
2005
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2006
{
2007
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2008
	struct queue_limits limits;
2009
	int r;
L
Linus Torvalds 已提交
2010

2011
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2012 2013

	/* device must be suspended */
2014
	if (!dm_suspended_md(md))
2015
		goto out;
L
Linus Torvalds 已提交
2016

2017 2018 2019 2020 2021 2022 2023
	/*
	 * 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
2024
		live_map = dm_get_live_table_fast(md);
2025 2026
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2027
		dm_put_live_table_fast(md);
2028 2029
	}

2030 2031 2032 2033 2034 2035
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2036
	}
2037

2038
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2039

2040
out:
2041
	mutex_unlock(&md->suspend_lock);
2042
	return map;
L
Linus Torvalds 已提交
2043 2044 2045 2046 2047 2048
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2049
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2050
{
2051
	int r;
L
Linus Torvalds 已提交
2052 2053

	WARN_ON(md->frozen_sb);
2054

2055
	md->frozen_sb = freeze_bdev(md->bdev);
2056
	if (IS_ERR(md->frozen_sb)) {
2057
		r = PTR_ERR(md->frozen_sb);
2058 2059
		md->frozen_sb = NULL;
		return r;
2060 2061
	}

2062 2063
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2064 2065 2066
	return 0;
}

2067
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2068
{
2069 2070 2071
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2072
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2073
	md->frozen_sb = NULL;
2074
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2075 2076 2077
}

/*
2078 2079 2080
 * 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.
2081
 *
2082
 * Caller must hold md->suspend_lock
2083
 */
2084
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
2085 2086
			unsigned suspend_flags, int interruptible,
			int dmf_suspended_flag)
L
Linus Torvalds 已提交
2087
{
2088 2089 2090
	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
2091

2092 2093 2094 2095 2096 2097 2098
	/*
	 * 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);

2099 2100 2101 2102
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
2103 2104
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2105
	/*
K
Kiyoshi Ueda 已提交
2106 2107 2108 2109
	 * 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
2110 2111 2112
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
2113 2114
		if (r) {
			dm_table_presuspend_undo_targets(map);
2115
			return r;
2116
		}
2117
	}
L
Linus Torvalds 已提交
2118 2119

	/*
2120 2121 2122 2123 2124 2125 2126
	 * 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
2127 2128 2129
	 * __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
2130
	 */
2131
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2132 2133
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2134

2135
	/*
2136 2137
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2138
	 */
2139
	if (dm_request_based(md)) {
2140
		dm_stop_queue(md->queue);
2141 2142
		if (md->kworker_task)
			flush_kthread_worker(&md->kworker);
2143
	}
2144

2145 2146
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2147
	/*
2148 2149 2150
	 * 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
2151
	 */
2152
	r = dm_wait_for_completion(md, interruptible);
2153 2154
	if (!r)
		set_bit(dmf_suspended_flag, &md->flags);
L
Linus Torvalds 已提交
2155

2156
	if (noflush)
2157
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2158 2159
	if (map)
		synchronize_srcu(&md->io_barrier);
2160

L
Linus Torvalds 已提交
2161
	/* were we interrupted ? */
2162
	if (r < 0) {
2163
		dm_queue_flush(md);
M
Milan Broz 已提交
2164

2165
		if (dm_request_based(md))
2166
			dm_start_queue(md->queue);
2167

2168
		unlock_fs(md);
2169
		dm_table_presuspend_undo_targets(map);
2170
		/* pushback list is already flushed, so skip flush */
2171
	}
L
Linus Torvalds 已提交
2172

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
	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;
	}

2214
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2215

2216
	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED);
2217 2218
	if (r)
		goto out_unlock;
2219

2220 2221
	dm_table_postsuspend_targets(map);

2222
out_unlock:
2223
	mutex_unlock(&md->suspend_lock);
2224
	return r;
L
Linus Torvalds 已提交
2225 2226
}

2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242
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))
2243
		dm_start_queue(md->queue);
2244 2245 2246 2247 2248 2249

	unlock_fs(md);

	return 0;
}

L
Linus Torvalds 已提交
2250 2251
int dm_resume(struct mapped_device *md)
{
2252 2253
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2254

2255 2256 2257
retry:
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

2258
	if (!dm_suspended_md(md))
2259 2260
		goto out;

2261 2262 2263 2264 2265 2266 2267 2268 2269
	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;
	}

2270
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2271
	if (!map || !dm_table_get_size(map))
2272
		goto out;
L
Linus Torvalds 已提交
2273

2274
	r = __dm_resume(md, map);
2275 2276
	if (r)
		goto out;
2277 2278 2279

	clear_bit(DMF_SUSPENDED, &md->flags);

2280 2281
	r = 0;
out:
2282
	mutex_unlock(&md->suspend_lock);
2283

2284
	return r;
L
Linus Torvalds 已提交
2285 2286
}

M
Mikulas Patocka 已提交
2287 2288 2289 2290 2291 2292
/*
 * 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.
 */

2293
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
2294
{
2295 2296
	struct dm_table *map = NULL;

2297
	if (md->internal_suspend_count++)
2298 2299 2300 2301 2302 2303 2304
		return; /* nested internal suspend */

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

2305
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2306 2307 2308 2309 2310 2311 2312

	/*
	 * 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.
	 */
2313 2314
	(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE,
			    DMF_SUSPENDED_INTERNALLY);
2315 2316 2317 2318 2319 2320

	dm_table_postsuspend_targets(map);
}

static void __dm_internal_resume(struct mapped_device *md)
{
2321 2322 2323
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
2324 2325
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
2326
	if (dm_suspended_md(md))
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
		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
2366 2367 2368 2369 2370 2371 2372
		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);
}
2373
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
2374

2375
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
2376
{
2377
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
2378 2379 2380 2381 2382 2383 2384
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
2385
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
2386

L
Linus Torvalds 已提交
2387 2388 2389
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2390
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2391
		       unsigned cookie)
2392
{
M
Milan Broz 已提交
2393 2394 2395 2396
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2397
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2398 2399 2400
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2401 2402
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2403
	}
2404 2405
}

M
Mike Anderson 已提交
2406 2407 2408 2409 2410
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421
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
2422 2423 2424 2425 2426 2427 2428 2429 2430
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
2431 2432 2433 2434 2435 2436 2437 2438
/*
 * 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;
}
2439
EXPORT_SYMBOL_GPL(dm_disk);
L
Linus Torvalds 已提交
2440

M
Milan Broz 已提交
2441 2442
struct kobject *dm_kobject(struct mapped_device *md)
{
2443
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
2444 2445 2446 2447 2448 2449
}

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

2450
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
2451

2452
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2453
	    dm_deleting_md(md))
2454 2455
		return NULL;

M
Milan Broz 已提交
2456 2457 2458 2459
	dm_get(md);
	return md;
}

2460
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2461 2462 2463 2464
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2465 2466 2467 2468 2469
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
2470 2471 2472 2473 2474
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

2475 2476
int dm_suspended(struct dm_target *ti)
{
2477
	return dm_suspended_md(dm_table_get_md(ti->table));
2478 2479 2480
}
EXPORT_SYMBOL_GPL(dm_suspended);

2481 2482
int dm_noflush_suspending(struct dm_target *ti)
{
2483
	return __noflush_suspending(dm_table_get_md(ti->table));
2484 2485 2486
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

2487
struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, unsigned type,
2488
					    unsigned integrity, unsigned per_io_data_size)
K
Kiyoshi Ueda 已提交
2489
{
2490
	struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id);
2491 2492
	struct kmem_cache *cachep = NULL;
	unsigned int pool_size = 0;
J
Jun'ichi Nomura 已提交
2493
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2494 2495

	if (!pools)
2496
		return NULL;
K
Kiyoshi Ueda 已提交
2497

2498 2499
	switch (type) {
	case DM_TYPE_BIO_BASED:
2500
	case DM_TYPE_DAX_BIO_BASED:
2501 2502
		cachep = _io_cache;
		pool_size = dm_get_reserved_bio_based_ios();
2503
		front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
		break;
	case DM_TYPE_REQUEST_BASED:
		cachep = _rq_tio_cache;
		pool_size = dm_get_reserved_rq_based_ios();
		pools->rq_pool = mempool_create_slab_pool(pool_size, _rq_cache);
		if (!pools->rq_pool)
			goto out;
		/* fall through to setup remaining rq-based pools */
	case DM_TYPE_MQ_REQUEST_BASED:
		if (!pool_size)
			pool_size = dm_get_reserved_rq_based_ios();
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
2516
		/* per_io_data_size is used for blk-mq pdu at queue allocation */
2517 2518 2519 2520 2521 2522 2523 2524 2525 2526
		break;
	default:
		BUG();
	}

	if (cachep) {
		pools->io_pool = mempool_create_slab_pool(pool_size, cachep);
		if (!pools->io_pool)
			goto out;
	}
K
Kiyoshi Ueda 已提交
2527

J
Junichi Nomura 已提交
2528
	pools->bs = bioset_create_nobvec(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
2529
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2530
		goto out;
K
Kiyoshi Ueda 已提交
2531

2532
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2533
		goto out;
2534

K
Kiyoshi Ueda 已提交
2535
	return pools;
2536 2537 2538

out:
	dm_free_md_mempools(pools);
2539

2540
	return NULL;
K
Kiyoshi Ueda 已提交
2541 2542 2543 2544 2545 2546 2547
}

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

2548 2549
	mempool_destroy(pools->io_pool);
	mempool_destroy(pools->rq_pool);
2550

K
Kiyoshi Ueda 已提交
2551 2552 2553 2554 2555 2556
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

2557 2558 2559 2560 2561 2562 2563 2564 2565
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)
2566 2567
{
	struct mapped_device *md = bdev->bd_disk->private_data;
2568 2569 2570
	struct dm_table *table;
	struct dm_target *ti;
	int ret = -ENOTTY, srcu_idx;
2571

2572 2573 2574
	table = dm_get_live_table(md, &srcu_idx);
	if (!table || !dm_table_get_size(table))
		goto out;
2575

2576 2577 2578 2579
	/* 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);
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 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
	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;
2627 2628 2629
}

static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2630
			 u32 flags)
2631 2632 2633 2634
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2635
	int r;
2636

2637
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2638 2639 2640 2641 2642 2643 2644 2645 2646
	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;

2647
	bdput(bdev);
2648 2649 2650 2651 2652 2653 2654 2655
	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;
2656
	int r;
2657

2658
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2659 2660 2661 2662 2663 2664 2665 2666 2667
	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;

2668
	bdput(bdev);
2669 2670 2671 2672
	return r;
}

static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2673
			 enum pr_type type, bool abort)
2674 2675 2676 2677
{
	struct mapped_device *md = bdev->bd_disk->private_data;
	const struct pr_ops *ops;
	fmode_t mode;
2678
	int r;
2679

2680
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2681 2682 2683 2684 2685 2686 2687 2688 2689
	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;

2690
	bdput(bdev);
2691 2692 2693 2694 2695 2696 2697 2698
	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;
2699
	int r;
2700

2701
	r = dm_grab_bdev_for_ioctl(md, &bdev, &mode);
2702 2703 2704 2705 2706 2707 2708 2709 2710
	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;

2711
	bdput(bdev);
2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
	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,
};

2723
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2724 2725
	.open = dm_blk_open,
	.release = dm_blk_close,
2726
	.ioctl = dm_blk_ioctl,
2727
	.direct_access = dm_blk_direct_access,
D
Darrick J. Wong 已提交
2728
	.getgeo = dm_blk_getgeo,
2729
	.pr_ops = &dm_pr_ops,
L
Linus Torvalds 已提交
2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740
	.owner = THIS_MODULE
};

/*
 * 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");
2741

2742 2743 2744
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

2745 2746 2747
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
2748 2749 2750
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");