dm.c 76.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 * Copyright (C) 2001, 2002 Sistina Software (UK) Limited.
M
Milan Broz 已提交
3
 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
L
Linus Torvalds 已提交
4 5 6 7 8
 *
 * This file is released under the GPL.
 */

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

#include <linux/init.h>
#include <linux/module.h>
A
Arjan van de Ven 已提交
13
#include <linux/mutex.h>
L
Linus Torvalds 已提交
14 15 16 17 18 19
#include <linux/moduleparam.h>
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/mempool.h>
#include <linux/slab.h>
#include <linux/idr.h>
D
Darrick J. Wong 已提交
20
#include <linux/hdreg.h>
21
#include <linux/delay.h>
22
#include <linux/wait.h>
23
#include <linux/kthread.h>
24 25

#include <trace/events/block.h>
L
Linus Torvalds 已提交
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
/*
K
Kiyoshi Ueda 已提交
62
 * For bio-based dm.
L
Linus Torvalds 已提交
63 64 65 66 67 68
 * 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
69
	struct bio *bio;
70
	unsigned long start_time;
71
	spinlock_t endio_lock;
M
Mikulas Patocka 已提交
72
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
73 74
};

K
Kiyoshi Ueda 已提交
75 76 77 78 79 80 81
/*
 * For request-based dm.
 * One of these is allocated per request.
 */
struct dm_rq_target_io {
	struct mapped_device *md;
	struct dm_target *ti;
82
	struct request *orig, *clone;
83
	struct kthread_work work;
K
Kiyoshi Ueda 已提交
84 85 86 87 88
	int error;
	union map_info info;
};

/*
89 90 91 92 93 94
 * For request-based dm - the bio clones we allocate are embedded in these
 * structs.
 *
 * We allocate these with bio_alloc_bioset, using the front_pad parameter when
 * the bioset is created - this means the bio has to come at the end of the
 * struct.
K
Kiyoshi Ueda 已提交
95 96 97
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
98
	struct dm_rq_target_io *tio;
99
	struct bio clone;
K
Kiyoshi Ueda 已提交
100 101
};

102 103 104 105 106 107 108 109
union map_info *dm_get_rq_mapinfo(struct request *rq)
{
	if (rq && rq->end_io_data)
		return &((struct dm_rq_target_io *)rq->end_io_data)->info;
	return NULL;
}
EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo);

110 111
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
112 113 114
/*
 * Bits for the md->flags field.
 */
115
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
116
#define DMF_SUSPENDED 1
117
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
118
#define DMF_FREEING 3
119
#define DMF_DELETING 4
120
#define DMF_NOFLUSH_SUSPENDING 5
121
#define DMF_MERGE_IS_OPTIONAL 6
M
Mikulas Patocka 已提交
122
#define DMF_DEFERRED_REMOVE 7
123
#define DMF_SUSPENDED_INTERNALLY 8
L
Linus Torvalds 已提交
124

M
Mikulas Patocka 已提交
125 126 127 128 129 130 131 132
/*
 * A dummy definition to make RCU happy.
 * struct dm_table should never be dereferenced in this file.
 */
struct dm_table {
	int undefined__;
};

133 134 135
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
136
struct mapped_device {
M
Mikulas Patocka 已提交
137
	struct srcu_struct io_barrier;
138
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
139
	atomic_t holders;
140
	atomic_t open_count;
L
Linus Torvalds 已提交
141

M
Mikulas Patocka 已提交
142 143 144 145 146
	/*
	 * The current mapping.
	 * Use dm_get_live_table{_fast} or take suspend_lock for
	 * dereference.
	 */
147
	struct dm_table __rcu *map;
M
Mikulas Patocka 已提交
148

149 150 151
	struct list_head table_devices;
	struct mutex table_devices_lock;

L
Linus Torvalds 已提交
152 153
	unsigned long flags;

154
	struct request_queue *queue;
155
	unsigned type;
156
	/* Protect queue and type against concurrent access. */
157 158
	struct mutex type_lock;

159 160
	struct target_type *immutable_target_type;

L
Linus Torvalds 已提交
161
	struct gendisk *disk;
M
Mike Anderson 已提交
162
	char name[16];
L
Linus Torvalds 已提交
163 164 165 166 167 168

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
169
	atomic_t pending[2];
L
Linus Torvalds 已提交
170
	wait_queue_head_t wait;
171
	struct work_struct work;
K
Kiyoshi Ueda 已提交
172
	struct bio_list deferred;
173
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
174

175
	/*
176
	 * Processing queue (flush)
177 178 179
	 */
	struct workqueue_struct *wq;

L
Linus Torvalds 已提交
180 181 182 183
	/*
	 * io objects are allocated from here.
	 */
	mempool_t *io_pool;
184
	mempool_t *rq_pool;
L
Linus Torvalds 已提交
185

S
Stefan Bader 已提交
186 187
	struct bio_set *bs;

L
Linus Torvalds 已提交
188 189 190 191 192
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
193 194 195
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
196 197 198 199 200

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
201
	struct block_device *bdev;
D
Darrick J. Wong 已提交
202 203 204

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
205

206 207
	/* kobject and completion */
	struct dm_kobject_holder kobj_holder;
208

209 210
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
M
Mikulas Patocka 已提交
211

212 213 214
	/* the number of internal suspends */
	unsigned internal_suspend_count;

M
Mikulas Patocka 已提交
215
	struct dm_stats stats;
216 217 218

	struct kthread_worker kworker;
	struct task_struct *kworker_task;
L
Linus Torvalds 已提交
219 220
};

K
Kiyoshi Ueda 已提交
221 222 223 224 225
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
226
	mempool_t *rq_pool;
K
Kiyoshi Ueda 已提交
227 228 229
	struct bio_set *bs;
};

230 231 232 233 234 235
struct table_device {
	struct list_head list;
	atomic_t count;
	struct dm_dev dm_dev;
};

236 237
#define RESERVED_BIO_BASED_IOS		16
#define RESERVED_REQUEST_BASED_IOS	256
238
#define RESERVED_MAX_IOS		1024
239
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
240
static struct kmem_cache *_rq_tio_cache;
241
static struct kmem_cache *_rq_cache;
242

243 244 245 246 247
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

248 249 250 251 252
/*
 * Request-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;

253
static unsigned __dm_get_module_param(unsigned *module_param,
254 255
				      unsigned def, unsigned max)
{
256 257
	unsigned param = ACCESS_ONCE(*module_param);
	unsigned modified_param = 0;
258

259 260 261 262
	if (!param)
		modified_param = def;
	else if (param > max)
		modified_param = max;
263

264 265 266
	if (modified_param) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
267 268
	}

269
	return param;
270 271
}

272 273
unsigned dm_get_reserved_bio_based_ios(void)
{
274
	return __dm_get_module_param(&reserved_bio_based_ios,
275 276 277 278
				     RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

279 280
unsigned dm_get_reserved_rq_based_ios(void)
{
281
	return __dm_get_module_param(&reserved_rq_based_ios,
282 283 284 285
				     RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);

L
Linus Torvalds 已提交
286 287
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
288
	int r = -ENOMEM;
L
Linus Torvalds 已提交
289 290

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
291
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
292
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
293
		return r;
L
Linus Torvalds 已提交
294

K
Kiyoshi Ueda 已提交
295 296
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
297
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
298

299 300 301 302 303
	_rq_cache = kmem_cache_create("dm_clone_request", sizeof(struct request),
				      __alignof__(struct request), 0, NULL);
	if (!_rq_cache)
		goto out_free_rq_tio_cache;

M
Mike Anderson 已提交
304
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
305
	if (r)
306
		goto out_free_rq_cache;
M
Mike Anderson 已提交
307

308 309 310 311 312 313
	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
314 315
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
316
	if (r < 0)
317
		goto out_free_workqueue;
L
Linus Torvalds 已提交
318 319 320 321 322

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
323

324 325
out_free_workqueue:
	destroy_workqueue(deferred_remove_workqueue);
K
Kiyoshi Ueda 已提交
326 327
out_uevent_exit:
	dm_uevent_exit();
328 329
out_free_rq_cache:
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
330 331
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
332 333 334 335
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
336 337 338 339
}

static void local_exit(void)
{
M
Mikulas Patocka 已提交
340
	flush_scheduled_work();
341
	destroy_workqueue(deferred_remove_workqueue);
M
Mikulas Patocka 已提交
342

343
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
344
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
345
	kmem_cache_destroy(_io_cache);
346
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
347
	dm_uevent_exit();
L
Linus Torvalds 已提交
348 349 350 351 352 353

	_major = 0;

	DMINFO("cleaned up");
}

354
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
355 356 357 358
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
359
	dm_io_init,
360
	dm_kcopyd_init,
L
Linus Torvalds 已提交
361
	dm_interface_init,
M
Mikulas Patocka 已提交
362
	dm_statistics_init,
L
Linus Torvalds 已提交
363 364
};

365
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
366 367 368 369
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
370
	dm_io_exit,
371
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
372
	dm_interface_exit,
M
Mikulas Patocka 已提交
373
	dm_statistics_exit,
L
Linus Torvalds 已提交
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402
};

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]();
403 404 405 406 407

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
408 409 410 411 412
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
413 414 415 416 417
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
418
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
419 420 421
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
422 423
	spin_lock(&_minor_lock);

A
Al Viro 已提交
424
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
425 426 427
	if (!md)
		goto out;

428
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
429
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
430 431 432 433
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
434
	dm_get(md);
435
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
436 437 438 439
out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
440 441
}

442
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
443
{
444
	struct mapped_device *md;
445

446 447
	spin_lock(&_minor_lock);

448 449 450 451
	md = disk->private_data;
	if (WARN_ON(!md))
		goto out;

M
Mikulas Patocka 已提交
452 453
	if (atomic_dec_and_test(&md->open_count) &&
	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
454
		queue_work(deferred_remove_workqueue, &deferred_remove_work);
M
Mikulas Patocka 已提交
455

L
Linus Torvalds 已提交
456
	dm_put(md);
457
out:
458
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
459 460
}

461 462 463 464 465 466 467 468
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
469
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
470 471 472 473 474
{
	int r = 0;

	spin_lock(&_minor_lock);

M
Mikulas Patocka 已提交
475
	if (dm_open_count(md)) {
476
		r = -EBUSY;
M
Mikulas Patocka 已提交
477 478 479 480
		if (mark_deferred)
			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
		r = -EEXIST;
481 482 483 484 485 486 487 488
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

M
Mikulas Patocka 已提交
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
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
510 511 512 513 514
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

515 516 517 518 519
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
	return md->queue;
}

M
Mikulas Patocka 已提交
520 521 522 523 524
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
525 526 527 528 529 530 531
static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
	struct mapped_device *md = bdev->bd_disk->private_data;

	return dm_get_geometry(md, geo);
}

A
Al Viro 已提交
532
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
533 534
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
535
	struct mapped_device *md = bdev->bd_disk->private_data;
M
Mikulas Patocka 已提交
536
	int srcu_idx;
537
	struct dm_table *map;
538 539 540
	struct dm_target *tgt;
	int r = -ENOTTY;

541
retry:
M
Mikulas Patocka 已提交
542 543
	map = dm_get_live_table(md, &srcu_idx);

544 545 546 547 548 549 550 551
	if (!map || !dm_table_get_size(map))
		goto out;

	/* We only support devices that have a single target */
	if (dm_table_get_num_targets(map) != 1)
		goto out;

	tgt = dm_table_get_target(map, 0);
552 553
	if (!tgt->type->ioctl)
		goto out;
554

555
	if (dm_suspended_md(md)) {
556 557 558 559
		r = -EAGAIN;
		goto out;
	}

560
	r = tgt->type->ioctl(tgt, cmd, arg);
561 562

out:
M
Mikulas Patocka 已提交
563
	dm_put_live_table(md, srcu_idx);
564

565 566 567 568 569
	if (r == -ENOTCONN) {
		msleep(10);
		goto retry;
	}

570 571 572
	return r;
}

A
Alasdair G Kergon 已提交
573
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
574 575 576 577
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
578
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
579 580 581 582
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
583
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
584
{
585
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
586 587
}

K
Kiyoshi Ueda 已提交
588 589
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
590
{
J
Jun'ichi Nomura 已提交
591
	return mempool_alloc(md->io_pool, gfp_mask);
592 593 594 595
}

static void free_rq_tio(struct dm_rq_target_io *tio)
{
J
Jun'ichi Nomura 已提交
596
	mempool_free(tio, tio->md->io_pool);
597 598
}

599 600 601 602 603 604 605 606 607 608 609
static struct request *alloc_clone_request(struct mapped_device *md,
					   gfp_t gfp_mask)
{
	return mempool_alloc(md->rq_pool, gfp_mask);
}

static void free_clone_request(struct mapped_device *md, struct request *rq)
{
	mempool_free(rq, md->rq_pool);
}

K
Kiyoshi Ueda 已提交
610 611 612 613 614 615
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

616 617 618
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
619
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
620
	int cpu;
M
Mikulas Patocka 已提交
621
	int rw = bio_data_dir(bio);
622 623 624

	io->start_time = jiffies;

T
Tejun Heo 已提交
625 626 627
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
628 629
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
630 631

	if (unlikely(dm_stats_used(&md->stats)))
632
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
M
Mikulas Patocka 已提交
633
				    bio_sectors(bio), false, 0, &io->stats_aux);
634 635
}

636
static void end_io_acct(struct dm_io *io)
637 638 639 640
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
641
	int pending;
642 643
	int rw = bio_data_dir(bio);

644
	generic_end_io_acct(rw, &dm_disk(md)->part0, io->start_time);
645

M
Mikulas Patocka 已提交
646
	if (unlikely(dm_stats_used(&md->stats)))
647
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
M
Mikulas Patocka 已提交
648 649
				    bio_sectors(bio), true, duration, &io->stats_aux);

650 651
	/*
	 * After this is decremented the bio must not be touched if it is
652
	 * a flush.
653
	 */
654 655
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
656
	pending += atomic_read(&md->pending[rw^0x1]);
657

658 659 660
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
661 662
}

L
Linus Torvalds 已提交
663 664 665
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
666
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
667
{
668
	unsigned long flags;
L
Linus Torvalds 已提交
669

670
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
671
	bio_list_add(&md->deferred, bio);
672
	spin_unlock_irqrestore(&md->deferred_lock, flags);
673
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
674 675 676 677 678
}

/*
 * 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
679
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
680
 */
M
Mikulas Patocka 已提交
681
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
682
{
M
Mikulas Patocka 已提交
683 684 685 686
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
687

M
Mikulas Patocka 已提交
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707
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
708

M
Mikulas Patocka 已提交
709 710 711
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
712 713
}

714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
/*
 * 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) {
		td = kmalloc(sizeof(*td), GFP_KERNEL);
		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
828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854
/*
 * 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
855 856 857 858 859 860 861 862 863
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

864 865 866 867 868
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
869 870 871 872
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
873
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
874
{
875
	unsigned long flags;
876 877 878
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
879 880

	/* Push-back supersedes any I/O errors */
881 882 883 884 885 886
	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
887 888

	if (atomic_dec_and_test(&io->io_count)) {
889 890 891 892
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
893
			spin_lock_irqsave(&md->deferred_lock, flags);
894 895 896
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
897 898
				/* noflush suspend was interrupted. */
				io->error = -EIO;
899
			spin_unlock_irqrestore(&md->deferred_lock, flags);
900 901
		}

902 903
		io_error = io->error;
		bio = io->bio;
904 905 906 907 908
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
909

910
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_iter.bi_size) {
911
			/*
912 913
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
914
			 */
915 916
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
917
		} else {
918
			/* done with normal IO or empty flush */
919
			trace_block_bio_complete(md->queue, bio, io_error);
920
			bio_endio(bio, io_error);
921
		}
L
Linus Torvalds 已提交
922 923 924
	}
}

925 926 927 928 929 930 931 932
static void disable_write_same(struct mapped_device *md)
{
	struct queue_limits *limits = dm_get_queue_limits(md);

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

933
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
934
{
935
	int r = error;
M
Mikulas Patocka 已提交
936
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
937
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
938
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
939 940 941 942 943 944
	dm_endio_fn endio = tio->ti->type->end_io;

	if (!bio_flagged(bio, BIO_UPTODATE) && !error)
		error = -EIO;

	if (endio) {
M
Mikulas Patocka 已提交
945
		r = endio(tio->ti, bio, error);
946 947 948 949 950
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
951
			error = r;
952 953
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
954
			return;
955 956 957 958
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
959 960
	}

961 962 963 964
	if (unlikely(r == -EREMOTEIO && (bio->bi_rw & REQ_WRITE_SAME) &&
		     !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors))
		disable_write_same(md);

S
Stefan Bader 已提交
965
	free_tio(md, tio);
966
	dec_pending(io, error);
L
Linus Torvalds 已提交
967 968
}

969 970 971 972 973
/*
 * Partial completion handling for request-based dm
 */
static void end_clone_bio(struct bio *clone, int error)
{
M
Mikulas Patocka 已提交
974 975
	struct dm_rq_clone_bio_info *info =
		container_of(clone, struct dm_rq_clone_bio_info, clone);
976 977
	struct dm_rq_target_io *tio = info->tio;
	struct bio *bio = info->orig;
978
	unsigned int nr_bytes = info->orig->bi_iter.bi_size;
979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024

	bio_put(clone);

	if (tio->error)
		/*
		 * An error has already been detected on the request.
		 * Once error occurred, just let clone->end_io() handle
		 * the remainder.
		 */
		return;
	else if (error) {
		/*
		 * Don't notice the error to the upper layer yet.
		 * The error handling decision is made by the target driver,
		 * when the request is completed.
		 */
		tio->error = error;
		return;
	}

	/*
	 * I/O for the bio successfully completed.
	 * Notice the data completion to the upper layer.
	 */

	/*
	 * bios are processed from the head of the list.
	 * So the completing bio should always be rq->bio.
	 * If it's not, something wrong is happening.
	 */
	if (tio->orig->bio != bio)
		DMERR("bio completion is going in the middle of the request");

	/*
	 * Update the original request.
	 * Do not use blk_end_request() here, because it may complete
	 * the original request before the clone, and break the ordering.
	 */
	blk_update_request(tio->orig, 0, nr_bytes);
}

/*
 * Don't touch any member of the md after calling this function because
 * the md may be freed in dm_put() at the end of this function.
 * Or do dm_get() before calling this function and dm_put() later.
 */
1025
static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
1026
{
1027 1028
	int nr_requests_pending;

1029
	atomic_dec(&md->pending[rw]);
1030 1031

	/* nudge anyone waiting on suspend queue */
1032 1033
	nr_requests_pending = md_in_flight(md);
	if (!nr_requests_pending)
1034 1035
		wake_up(&md->wait);

1036 1037 1038 1039 1040 1041
	/*
	 * Run this off this callpath, as drivers could invoke end_io while
	 * inside their request_fn (and holding the queue lock). Calling
	 * back into ->request_fn() could deadlock attempting to grab the
	 * queue lock again.
	 */
1042 1043 1044 1045 1046
	if (run_queue) {
		if (!nr_requests_pending ||
		    (nr_requests_pending >= md->queue->nr_congestion_on))
			blk_run_queue_async(md->queue);
	}
1047 1048 1049 1050 1051 1052 1053

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

1054 1055 1056 1057 1058
static void free_rq_clone(struct request *clone)
{
	struct dm_rq_target_io *tio = clone->end_io_data;

	blk_rq_unprep_clone(clone);
1059 1060 1061 1062
	if (clone->q && clone->q->mq_ops)
		tio->ti->type->release_clone_rq(clone);
	else
		free_clone_request(tio->md, clone);
1063 1064 1065
	free_rq_tio(tio);
}

K
Kiyoshi Ueda 已提交
1066 1067
/*
 * Complete the clone and the original request.
1068 1069
 * Must be called without clone's queue lock held,
 * see end_clone_request() for more details.
K
Kiyoshi Ueda 已提交
1070 1071 1072 1073 1074 1075 1076 1077
 */
static void dm_end_request(struct request *clone, int error)
{
	int rw = rq_data_dir(clone);
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;

1078
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
		rq->errors = clone->errors;
		rq->resid_len = clone->resid_len;

		if (rq->sense)
			/*
			 * We are using the sense buffer of the original
			 * request.
			 * So setting the length of the sense data is enough.
			 */
			rq->sense_len = clone->sense_len;
	}

	free_rq_clone(clone);
1092 1093
	blk_end_request_all(rq, error);
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
1094 1095
}

1096 1097
static void dm_unprep_request(struct request *rq)
{
1098 1099
	struct dm_rq_target_io *tio = rq->special;
	struct request *clone = tio->clone;
1100 1101 1102 1103

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

1104 1105
	if (clone)
		free_rq_clone(clone);
1106 1107 1108 1109 1110
}

/*
 * Requeue the original request of a clone.
 */
1111 1112
static void dm_requeue_unmapped_original_request(struct mapped_device *md,
						 struct request *rq)
1113
{
1114
	int rw = rq_data_dir(rq);
1115 1116 1117 1118 1119 1120 1121 1122 1123
	struct request_queue *q = rq->q;
	unsigned long flags;

	dm_unprep_request(rq);

	spin_lock_irqsave(q->queue_lock, flags);
	blk_requeue_request(q, rq);
	spin_unlock_irqrestore(q->queue_lock, flags);

1124 1125 1126 1127 1128 1129 1130 1131
	rq_completed(md, rw, false);
}

static void dm_requeue_unmapped_request(struct request *clone)
{
	struct dm_rq_target_io *tio = clone->end_io_data;

	dm_requeue_unmapped_original_request(tio->md, tio->orig);
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
}

static void __stop_queue(struct request_queue *q)
{
	blk_stop_queue(q);
}

static void stop_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	__stop_queue(q);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

static void __start_queue(struct request_queue *q)
{
	if (blk_queue_stopped(q))
		blk_start_queue(q);
}

static void start_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	__start_queue(q);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

1163
static void dm_done(struct request *clone, int error, bool mapped)
1164
{
1165
	int r = error;
1166
	struct dm_rq_target_io *tio = clone->end_io_data;
1167
	dm_request_endio_fn rq_end_io = NULL;
1168

1169 1170 1171 1172 1173 1174
	if (tio->ti) {
		rq_end_io = tio->ti->type->rq_end_io;

		if (mapped && rq_end_io)
			r = rq_end_io(tio->ti, clone, error, &tio->info);
	}
1175

1176 1177 1178 1179
	if (unlikely(r == -EREMOTEIO && (clone->cmd_flags & REQ_WRITE_SAME) &&
		     !clone->q->limits.max_write_same_sectors))
		disable_write_same(tio->md);

1180
	if (r <= 0)
1181
		/* The target wants to complete the I/O */
1182 1183
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
1184 1185
		/* The target will handle the I/O */
		return;
1186
	else if (r == DM_ENDIO_REQUEUE)
1187 1188 1189
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
1190
		DMWARN("unimplemented target endio return value: %d", r);
1191 1192 1193 1194
		BUG();
	}
}

1195 1196 1197 1198 1199 1200
/*
 * Request completion handler for request-based dm
 */
static void dm_softirq_done(struct request *rq)
{
	bool mapped = true;
1201 1202
	struct dm_rq_target_io *tio = rq->special;
	struct request *clone = tio->clone;
1203

1204 1205 1206 1207 1208 1209
	if (!clone) {
		blk_end_request_all(rq, tio->error);
		rq_completed(tio->md, rq_data_dir(rq), false);
		free_rq_tio(tio);
		return;
	}
1210 1211 1212 1213 1214 1215 1216

	if (rq->cmd_flags & REQ_FAILED)
		mapped = false;

	dm_done(clone, tio->error, mapped);
}

1217 1218 1219 1220
/*
 * Complete the clone and the original request with the error status
 * through softirq context.
 */
1221
static void dm_complete_request(struct request *rq, int error)
1222
{
1223
	struct dm_rq_target_io *tio = rq->special;
1224 1225 1226 1227 1228 1229 1230 1231 1232

	tio->error = error;
	blk_complete_request(rq);
}

/*
 * Complete the not-mapped clone and the original request with the error status
 * through softirq context.
 * Target's rq_end_io() function isn't called.
1233
 * This may be used when the target's map_rq() or clone_and_map_rq() functions fail.
1234
 */
1235
static void dm_kill_unmapped_request(struct request *rq, int error)
1236 1237
{
	rq->cmd_flags |= REQ_FAILED;
1238
	dm_complete_request(rq, error);
1239 1240 1241
}

/*
1242
 * Called with the clone's queue lock held
1243 1244 1245
 */
static void end_clone_request(struct request *clone, int error)
{
1246 1247
	struct dm_rq_target_io *tio = clone->end_io_data;

1248 1249 1250 1251 1252 1253 1254 1255 1256
	if (!clone->q->mq_ops) {
		/*
		 * For just cleaning up the information of the queue in which
		 * the clone was dispatched.
		 * The clone is *NOT* freed actually here because it is alloced
		 * from dm own mempool (REQ_ALLOCED isn't set).
		 */
		__blk_put_request(clone->q, clone);
	}
1257 1258 1259

	/*
	 * Actual request completion is done in a softirq context which doesn't
1260
	 * hold the clone's queue lock.  Otherwise, deadlock could occur because:
1261 1262 1263
	 *     - another request may be submitted by the upper level driver
	 *       of the stacking during the completion
	 *     - the submission which requires queue lock may be done
1264
	 *       against this clone's queue
1265
	 */
1266
	dm_complete_request(tio->orig, error);
1267 1268
}

1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
/*
 * 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
1281
{
1282
	sector_t len = max_io_len_target_boundary(sector, ti);
1283
	sector_t offset, max_len;
L
Linus Torvalds 已提交
1284 1285

	/*
1286
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
1287
	 */
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
	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
1298 1299 1300 1301 1302
	}

	return len;
}

1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
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);

1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
/*
 * A target may call dm_accept_partial_bio only from the map routine.  It is
 * allowed for all bio types except REQ_FLUSH.
 *
 * 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;
	BUG_ON(bio->bi_rw & REQ_FLUSH);
	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
1358
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1359 1360
{
	int r;
1361
	sector_t sector;
S
Stefan Bader 已提交
1362
	struct mapped_device *md;
1363
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1364
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1365 1366 1367 1368 1369 1370 1371 1372 1373

	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);
1374
	sector = clone->bi_iter.bi_sector;
M
Mikulas Patocka 已提交
1375
	r = ti->type->map(ti, clone);
1376
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1377
		/* the bio has been remapped so dispatch it */
1378

1379 1380
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
1381

L
Linus Torvalds 已提交
1382
		generic_make_request(clone);
1383 1384
	} 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
1385 1386 1387
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1388 1389 1390
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1400
	unsigned sector_count;
L
Linus Torvalds 已提交
1401 1402
};

1403
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1404
{
1405 1406
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1407 1408 1409 1410 1411
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1412
static void clone_bio(struct dm_target_io *tio, struct bio *bio,
1413
		      sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1414
{
1415
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1416

1417 1418 1419 1420
	__bio_clone_fast(clone, bio);

	if (bio_integrity(bio))
		bio_integrity_clone(clone, bio, GFP_NOIO);
A
Alasdair G Kergon 已提交
1421

1422 1423 1424 1425 1426
	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);
L
Linus Torvalds 已提交
1427 1428
}

1429
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1430
				      struct dm_target *ti,
1431
				      unsigned target_bio_nr)
1432
{
1433 1434 1435
	struct dm_target_io *tio;
	struct bio *clone;

1436
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1437
	tio = container_of(clone, struct dm_target_io, clone);
1438 1439 1440

	tio->io = ci->io;
	tio->ti = ti;
1441
	tio->target_bio_nr = target_bio_nr;
1442 1443 1444 1445

	return tio;
}

1446 1447
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1448
				       unsigned target_bio_nr, unsigned *len)
1449
{
1450
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1451
	struct bio *clone = &tio->clone;
1452

1453 1454
	tio->len_ptr = len;

1455
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1456
	if (len)
1457
		bio_setup_sector(clone, ci->sector, *len);
1458

A
Alasdair G Kergon 已提交
1459
	__map_bio(tio);
1460 1461
}

1462
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1463
				  unsigned num_bios, unsigned *len)
1464
{
1465
	unsigned target_bio_nr;
1466

1467
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1468
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1469 1470
}

1471
static int __send_empty_flush(struct clone_info *ci)
1472
{
1473
	unsigned target_nr = 0;
1474 1475
	struct dm_target *ti;

1476
	BUG_ON(bio_has_data(ci->bio));
1477
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1478
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1479 1480 1481 1482

	return 0;
}

A
Alasdair G Kergon 已提交
1483
static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1484
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1485
{
1486
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1487
	struct dm_target_io *tio;
1488 1489
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
M
Mike Snitzer 已提交
1490

1491 1492 1493 1494 1495
	/*
	 * 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
1496

1497
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1498
		tio = alloc_tio(ci, ti, target_bio_nr);
1499 1500
		tio->len_ptr = len;
		clone_bio(tio, bio, sector, *len);
1501 1502
		__map_bio(tio);
	}
M
Mike Snitzer 已提交
1503 1504
}

1505
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1506

1507
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1508
{
1509
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1510 1511
}

1512
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1513
{
1514
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1515 1516 1517 1518 1519 1520
}

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

static bool is_split_required_for_discard(struct dm_target *ti)
{
1521
	return ti->split_discard_bios;
M
Mike Snitzer 已提交
1522 1523
}

1524 1525 1526
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
M
Mike Snitzer 已提交
1527 1528
{
	struct dm_target *ti;
1529
	unsigned len;
1530
	unsigned num_bios;
M
Mike Snitzer 已提交
1531

1532 1533 1534 1535
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
M
Mike Snitzer 已提交
1536 1537

		/*
M
Mike Snitzer 已提交
1538 1539
		 * 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
1540
		 * reconfiguration might also have changed that since the
1541
		 * check was performed.
M
Mike Snitzer 已提交
1542
		 */
1543 1544
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1545
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1546

M
Mike Snitzer 已提交
1547
		if (is_split_required && !is_split_required(ti))
1548
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1549
		else
1550
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1551

1552
		__send_duplicate_bios(ci, ti, num_bios, &len);
1553 1554 1555

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1556 1557 1558 1559

	return 0;
}

1560
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1561
{
1562 1563
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1564 1565
}

1566
static int __send_write_same(struct clone_info *ci)
M
Mike Snitzer 已提交
1567
{
1568
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
M
Mike Snitzer 已提交
1569 1570
}

A
Alasdair G Kergon 已提交
1571 1572 1573
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1574
static int __split_and_process_non_flush(struct clone_info *ci)
L
Linus Torvalds 已提交
1575
{
1576
	struct bio *bio = ci->bio;
1577
	struct dm_target *ti;
1578
	unsigned len;
L
Linus Torvalds 已提交
1579

M
Mike Snitzer 已提交
1580
	if (unlikely(bio->bi_rw & REQ_DISCARD))
1581
		return __send_discard(ci);
M
Mike Snitzer 已提交
1582
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
1583
		return __send_write_same(ci);
M
Mike Snitzer 已提交
1584

1585 1586 1587 1588
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1589
	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
L
Linus Torvalds 已提交
1590

1591
	__clone_and_map_data_bio(ci, ti, ci->sector, &len);
L
Linus Torvalds 已提交
1592

1593 1594
	ci->sector += len;
	ci->sector_count -= len;
L
Linus Torvalds 已提交
1595

1596
	return 0;
L
Linus Torvalds 已提交
1597 1598 1599
}

/*
1600
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1601
 */
M
Mikulas Patocka 已提交
1602 1603
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
L
Linus Torvalds 已提交
1604 1605
{
	struct clone_info ci;
1606
	int error = 0;
L
Linus Torvalds 已提交
1607

M
Mikulas Patocka 已提交
1608
	if (unlikely(!map)) {
1609
		bio_io_error(bio);
1610 1611
		return;
	}
1612

M
Mikulas Patocka 已提交
1613
	ci.map = map;
L
Linus Torvalds 已提交
1614 1615 1616 1617 1618 1619
	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;
1620
	spin_lock_init(&ci.io->endio_lock);
1621
	ci.sector = bio->bi_iter.bi_sector;
L
Linus Torvalds 已提交
1622

1623
	start_io_acct(ci.io);
A
Alasdair G Kergon 已提交
1624

1625 1626 1627
	if (bio->bi_rw & REQ_FLUSH) {
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1628
		error = __send_empty_flush(&ci);
1629 1630
		/* dec_pending submits any data associated with flush */
	} else {
1631
		ci.bio = bio;
1632
		ci.sector_count = bio_sectors(bio);
1633
		while (ci.sector_count && !error)
1634
			error = __split_and_process_non_flush(&ci);
1635
	}
L
Linus Torvalds 已提交
1636 1637

	/* drop the extra reference count */
1638
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1639 1640 1641 1642 1643
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1644 1645 1646 1647 1648
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1649
	struct dm_table *map = dm_get_live_table_fast(md);
M
Milan Broz 已提交
1650 1651
	struct dm_target *ti;
	sector_t max_sectors;
1652
	int max_size = 0;
M
Milan Broz 已提交
1653 1654

	if (unlikely(!map))
1655
		goto out;
M
Milan Broz 已提交
1656 1657

	ti = dm_table_find_target(map, bvm->bi_sector);
1658
	if (!dm_target_is_valid(ti))
M
Mikulas Patocka 已提交
1659
		goto out;
M
Milan Broz 已提交
1660 1661 1662 1663

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1664
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
1665
			  (sector_t) queue_max_sectors(q));
M
Milan Broz 已提交
1666
	max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size;
1667
	if (unlikely(max_size < 0)) /* this shouldn't _ever_ happen */
M
Milan Broz 已提交
1668 1669 1670 1671 1672 1673 1674 1675 1676
		max_size = 0;

	/*
	 * merge_bvec_fn() returns number of bytes
	 * it can accept at this offset
	 * max is precomputed maximal io size
	 */
	if (max_size && ti->type->merge)
		max_size = ti->type->merge(ti, bvm, biovec, max_size);
1677 1678
	/*
	 * If the target doesn't support merge method and some of the devices
1679 1680 1681 1682
	 * provided their merge_bvec method (we know this by looking for the
	 * max_hw_sectors that dm_set_device_limits may set), then we can't
	 * allow bios with multiple vector entries.  So always set max_size
	 * to 0, and the code below allows just one page.
1683 1684 1685
	 */
	else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9)
		max_size = 0;
M
Milan Broz 已提交
1686

1687
out:
M
Mikulas Patocka 已提交
1688
	dm_put_live_table_fast(md);
M
Milan Broz 已提交
1689 1690 1691 1692 1693 1694 1695 1696 1697
	/*
	 * Always allow an entire first page
	 */
	if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT))
		max_size = biovec->bv_len;

	return max_size;
}

L
Linus Torvalds 已提交
1698 1699 1700 1701
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1702
static void dm_make_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1703
{
1704
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1705
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1706 1707
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
1708

M
Mikulas Patocka 已提交
1709
	map = dm_get_live_table(md, &srcu_idx);
L
Linus Torvalds 已提交
1710

1711
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1712

1713 1714
	/* 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
1715
		dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
1716

1717 1718 1719
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1720
			bio_io_error(bio);
1721
		return;
L
Linus Torvalds 已提交
1722 1723
	}

M
Mikulas Patocka 已提交
1724 1725
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1726
	return;
1727 1728
}

M
Mikulas Patocka 已提交
1729
int dm_request_based(struct mapped_device *md)
1730 1731 1732 1733
{
	return blk_queue_stackable(md->queue);
}

1734
static void dm_dispatch_clone_request(struct request *clone, struct request *rq)
1735 1736 1737
{
	int r;

1738 1739
	if (blk_queue_io_stat(clone->q))
		clone->cmd_flags |= REQ_IO_STAT;
1740

1741 1742
	clone->start_time = jiffies;
	r = blk_insert_cloned_request(clone->q, clone);
1743
	if (r)
1744
		/* must complete clone in terms of original request */
1745 1746 1747 1748 1749 1750 1751
		dm_complete_request(rq, r);
}

static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
				 void *data)
{
	struct dm_rq_target_io *tio = data;
1752 1753
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);
1754 1755 1756 1757 1758 1759 1760 1761 1762

	info->orig = bio_orig;
	info->tio = tio;
	bio->bi_end_io = end_clone_bio;

	return 0;
}

static int setup_clone(struct request *clone, struct request *rq,
1763
		       struct dm_rq_target_io *tio, gfp_t gfp_mask)
1764
{
1765
	int r;
1766

1767
	r = blk_rq_prep_clone(clone, rq, tio->md->bs, gfp_mask,
1768 1769 1770
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;
1771

1772 1773 1774
	clone->cmd = rq->cmd;
	clone->cmd_len = rq->cmd_len;
	clone->sense = rq->sense;
1775 1776 1777
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

1778 1779
	tio->clone = clone;

1780 1781 1782
	return 0;
}

K
Kiyoshi Ueda 已提交
1783
static struct request *clone_rq(struct request *rq, struct mapped_device *md,
1784
				struct dm_rq_target_io *tio, gfp_t gfp_mask)
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
{
	struct request *clone = alloc_clone_request(md, gfp_mask);

	if (!clone)
		return NULL;

	blk_rq_init(NULL, clone);
	if (setup_clone(clone, rq, tio, gfp_mask)) {
		/* -ENOMEM */
		free_clone_request(md, clone);
		return NULL;
	}

	return clone;
}

1801 1802
static void map_tio_request(struct kthread_work *work);

1803 1804
static struct dm_rq_target_io *prep_tio(struct request *rq,
					struct mapped_device *md, gfp_t gfp_mask)
K
Kiyoshi Ueda 已提交
1805 1806
{
	struct dm_rq_target_io *tio;
1807 1808
	int srcu_idx;
	struct dm_table *table;
K
Kiyoshi Ueda 已提交
1809 1810 1811 1812 1813 1814 1815

	tio = alloc_rq_tio(md, gfp_mask);
	if (!tio)
		return NULL;

	tio->md = md;
	tio->ti = NULL;
1816
	tio->clone = NULL;
K
Kiyoshi Ueda 已提交
1817 1818 1819
	tio->orig = rq;
	tio->error = 0;
	memset(&tio->info, 0, sizeof(tio->info));
1820
	init_kthread_work(&tio->work, map_tio_request);
K
Kiyoshi Ueda 已提交
1821

1822 1823 1824 1825 1826 1827 1828
	table = dm_get_live_table(md, &srcu_idx);
	if (!dm_table_mq_request_based(table)) {
		if (!clone_rq(rq, md, tio, gfp_mask)) {
			dm_put_live_table(md, srcu_idx);
			free_rq_tio(tio);
			return NULL;
		}
K
Kiyoshi Ueda 已提交
1829
	}
1830
	dm_put_live_table(md, srcu_idx);
K
Kiyoshi Ueda 已提交
1831

1832
	return tio;
K
Kiyoshi Ueda 已提交
1833 1834
}

1835 1836 1837 1838 1839 1840
/*
 * Called with the queue lock held.
 */
static int dm_prep_fn(struct request_queue *q, struct request *rq)
{
	struct mapped_device *md = q->queuedata;
1841
	struct dm_rq_target_io *tio;
1842 1843 1844 1845 1846 1847

	if (unlikely(rq->special)) {
		DMWARN("Already has something in rq->special.");
		return BLKPREP_KILL;
	}

1848 1849
	tio = prep_tio(rq, md, GFP_ATOMIC);
	if (!tio)
1850 1851
		return BLKPREP_DEFER;

1852
	rq->special = tio;
1853 1854 1855 1856 1857
	rq->cmd_flags |= REQ_DONTPREP;

	return BLKPREP_OK;
}

1858 1859
/*
 * Returns:
1860 1861 1862
 * 0                : the request has been processed
 * DM_MAPIO_REQUEUE : the original request needs to be requeued
 * < 0              : the request was completed due to failure
1863
 */
1864
static int map_request(struct dm_target *ti, struct request *rq,
1865
		       struct mapped_device *md)
1866
{
1867
	int r;
1868
	struct dm_rq_target_io *tio = rq->special;
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
	struct request *clone = NULL;

	if (tio->clone) {
		clone = tio->clone;
		r = ti->type->map_rq(ti, clone, &tio->info);
	} else {
		r = ti->type->clone_and_map_rq(ti, rq, &tio->info, &clone);
		if (r < 0) {
			/* The target wants to complete the I/O */
			dm_kill_unmapped_request(rq, r);
			return r;
		}
		if (IS_ERR(clone))
			return DM_MAPIO_REQUEUE;
		if (setup_clone(clone, rq, tio, GFP_KERNEL)) {
			/* -ENOMEM */
			ti->type->release_clone_rq(clone);
			return DM_MAPIO_REQUEUE;
		}
	}
1889 1890 1891 1892 1893 1894 1895

	switch (r) {
	case DM_MAPIO_SUBMITTED:
		/* The target has taken the I/O to submit by itself later */
		break;
	case DM_MAPIO_REMAPPED:
		/* The target has remapped the I/O so dispatch it */
1896
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
1897 1898
				     blk_rq_pos(rq));
		dm_dispatch_clone_request(clone, rq);
1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
		break;
	default:
		if (r > 0) {
			DMWARN("unimplemented target map return value: %d", r);
			BUG();
		}

		/* The target wants to complete the I/O */
1911
		dm_kill_unmapped_request(rq, r);
1912
		return r;
1913
	}
1914

1915
	return 0;
1916 1917
}

1918
static void map_tio_request(struct kthread_work *work)
1919
{
1920
	struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work);
1921 1922
	struct request *rq = tio->orig;
	struct mapped_device *md = tio->md;
1923

1924 1925
	if (map_request(tio->ti, rq, md) == DM_MAPIO_REQUEUE)
		dm_requeue_unmapped_original_request(md, rq);
1926 1927
}

1928
static void dm_start_request(struct mapped_device *md, struct request *orig)
1929 1930
{
	blk_start_request(orig);
1931
	atomic_inc(&md->pending[rq_data_dir(orig)]);
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942

	/*
	 * Hold the md reference here for the in-flight I/O.
	 * We can't rely on the reference count by device opener,
	 * because the device may be closed during the request completion
	 * when all bios are completed.
	 * See the comment in rq_completed() too.
	 */
	dm_get(md);
}

1943 1944 1945 1946 1947 1948 1949
/*
 * q->request_fn for request-based dm.
 * Called with the queue lock held.
 */
static void dm_request_fn(struct request_queue *q)
{
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1950 1951
	int srcu_idx;
	struct dm_table *map = dm_get_live_table(md, &srcu_idx);
1952
	struct dm_target *ti;
1953
	struct request *rq;
1954
	struct dm_rq_target_io *tio;
1955
	sector_t pos;
1956 1957

	/*
1958 1959 1960 1961
	 * For suspend, check blk_queue_stopped() and increment
	 * ->pending within a single queue_lock not to increment the
	 * number of in-flight I/Os after the queue is stopped in
	 * dm_suspend().
1962
	 */
J
Jens Axboe 已提交
1963
	while (!blk_queue_stopped(q)) {
1964 1965
		rq = blk_peek_request(q);
		if (!rq)
1966
			goto out;
1967

1968 1969 1970 1971 1972 1973
		/* always use block 0 to find the target for flushes for now */
		pos = 0;
		if (!(rq->cmd_flags & REQ_FLUSH))
			pos = blk_rq_pos(rq);

		ti = dm_table_find_target(map, pos);
1974 1975
		if (!dm_target_is_valid(ti)) {
			/*
1976
			 * Must perform setup, that rq_completed() requires,
1977 1978 1979
			 * before calling dm_kill_unmapped_request
			 */
			DMERR_LIMIT("request attempted access beyond the end of device");
1980 1981
			dm_start_request(md, rq);
			dm_kill_unmapped_request(rq, -EIO);
1982 1983
			continue;
		}
1984

1985
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
1986
			goto delay_and_out;
1987

1988
		dm_start_request(md, rq);
1989

1990 1991 1992 1993
		tio = rq->special;
		/* Establish tio->ti before queuing work (map_tio_request) */
		tio->ti = ti;
		queue_kthread_work(&md->kworker, &tio->work);
1994
		BUG_ON(!irqs_disabled());
1995 1996 1997 1998
	}

	goto out;

J
Jens Axboe 已提交
1999
delay_and_out:
2000
	blk_delay_queue(q, HZ / 100);
2001
out:
M
Mikulas Patocka 已提交
2002
	dm_put_live_table(md, srcu_idx);
2003 2004
}

L
Linus Torvalds 已提交
2005 2006
static int dm_any_congested(void *congested_data, int bdi_bits)
{
2007 2008 2009
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
2010

2011
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mikulas Patocka 已提交
2012
		map = dm_get_live_table_fast(md);
2013
		if (map) {
2014 2015 2016 2017 2018 2019 2020 2021 2022
			/*
			 * Request-based dm cares about only own queue for
			 * the query about congestion status of request_queue
			 */
			if (dm_request_based(md))
				r = md->queue->backing_dev_info.state &
				    bdi_bits;
			else
				r = dm_table_any_congested(map, bdi_bits);
2023
		}
M
Mikulas Patocka 已提交
2024
		dm_put_live_table_fast(md);
2025 2026
	}

L
Linus Torvalds 已提交
2027 2028 2029 2030 2031 2032
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
2033
static void free_minor(int minor)
L
Linus Torvalds 已提交
2034
{
2035
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2036
	idr_remove(&_minor_idr, minor);
2037
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2038 2039 2040 2041 2042
}

/*
 * See if the device with a specific minor # is free.
 */
2043
static int specific_minor(int minor)
L
Linus Torvalds 已提交
2044
{
T
Tejun Heo 已提交
2045
	int r;
L
Linus Torvalds 已提交
2046 2047 2048 2049

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

T
Tejun Heo 已提交
2050
	idr_preload(GFP_KERNEL);
2051
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2052

T
Tejun Heo 已提交
2053
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
2054

2055
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
2056 2057 2058 2059
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
2060 2061
}

2062
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
2063
{
T
Tejun Heo 已提交
2064
	int r;
J
Jeff Mahoney 已提交
2065

T
Tejun Heo 已提交
2066
	idr_preload(GFP_KERNEL);
2067
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2068

T
Tejun Heo 已提交
2069
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
2070

2071
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
2072 2073 2074 2075 2076
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
2077 2078
}

2079
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
2080

2081 2082
static void dm_wq_work(struct work_struct *work);

2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098
static void dm_init_md_queue(struct mapped_device *md)
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
	 * devices.  The type of this dm device has not been decided yet.
	 * The type is decided at the first table loading time.
	 * To prevent problematic device stacking, clear the queue flag
	 * for request stacking support until then.
	 *
	 * This queue is new, so no concurrency on the queue_flags.
	 */
	queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue);

	md->queue->queuedata = md;
	md->queue->backing_dev_info.congested_fn = dm_any_congested;
	md->queue->backing_dev_info.congested_data = md;
2099

2100 2101 2102
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

L
Linus Torvalds 已提交
2103 2104 2105
/*
 * Allocate and initialise a blank device with a given minor.
 */
2106
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
2107 2108
{
	int r;
2109
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
2110
	void *old_md;
L
Linus Torvalds 已提交
2111 2112 2113 2114 2115 2116

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

2117
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
2118
		goto bad_module_get;
2119

L
Linus Torvalds 已提交
2120
	/* get a minor number for the dev */
2121
	if (minor == DM_ANY_MINOR)
2122
		r = next_free_minor(&minor);
2123
	else
2124
		r = specific_minor(minor);
L
Linus Torvalds 已提交
2125
	if (r < 0)
M
Milan Broz 已提交
2126
		goto bad_minor;
L
Linus Torvalds 已提交
2127

M
Mikulas Patocka 已提交
2128 2129 2130 2131
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

2132
	md->type = DM_TYPE_NONE;
2133
	mutex_init(&md->suspend_lock);
2134
	mutex_init(&md->type_lock);
2135
	mutex_init(&md->table_devices_lock);
2136
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
2137
	atomic_set(&md->holders, 1);
2138
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
2139
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
2140 2141
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
2142
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
2143
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
2144

2145
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
2146
	if (!md->queue)
M
Milan Broz 已提交
2147
		goto bad_queue;
L
Linus Torvalds 已提交
2148

2149
	dm_init_md_queue(md);
S
Stefan Bader 已提交
2150

L
Linus Torvalds 已提交
2151 2152
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
2153
		goto bad_disk;
L
Linus Torvalds 已提交
2154

2155 2156
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
2157
	init_waitqueue_head(&md->wait);
2158
	INIT_WORK(&md->work, dm_wq_work);
2159
	init_waitqueue_head(&md->eventq);
2160
	init_completion(&md->kobj_holder.completion);
2161
	md->kworker_task = NULL;
2162

L
Linus Torvalds 已提交
2163 2164 2165 2166 2167 2168 2169
	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
2170
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
2171

T
Tejun Heo 已提交
2172
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
2173 2174 2175
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
2176 2177 2178 2179
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

2180 2181 2182 2183
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
	md->flush_bio.bi_rw = WRITE_FLUSH;

M
Mikulas Patocka 已提交
2184 2185
	dm_stats_init(&md->stats);

2186
	/* Populate the mapping, nobody knows we exist yet */
2187
	spin_lock(&_minor_lock);
2188
	old_md = idr_replace(&_minor_idr, md, minor);
2189
	spin_unlock(&_minor_lock);
2190 2191 2192

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
2193 2194
	return md;

M
Mikulas Patocka 已提交
2195 2196
bad_bdev:
	destroy_workqueue(md->wq);
2197
bad_thread:
2198
	del_gendisk(md->disk);
2199
	put_disk(md->disk);
M
Milan Broz 已提交
2200
bad_disk:
2201
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
2202
bad_queue:
M
Mikulas Patocka 已提交
2203 2204
	cleanup_srcu_struct(&md->io_barrier);
bad_io_barrier:
L
Linus Torvalds 已提交
2205
	free_minor(minor);
M
Milan Broz 已提交
2206
bad_minor:
2207
	module_put(THIS_MODULE);
M
Milan Broz 已提交
2208
bad_module_get:
L
Linus Torvalds 已提交
2209 2210 2211 2212
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
2213 2214
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
2215 2216
static void free_dev(struct mapped_device *md)
{
2217
	int minor = MINOR(disk_devt(md->disk));
2218

M
Mikulas Patocka 已提交
2219
	unlock_fs(md);
2220
	destroy_workqueue(md->wq);
2221 2222 2223

	if (md->kworker_task)
		kthread_stop(md->kworker_task);
K
Kiyoshi Ueda 已提交
2224 2225
	if (md->io_pool)
		mempool_destroy(md->io_pool);
2226 2227
	if (md->rq_pool)
		mempool_destroy(md->rq_pool);
K
Kiyoshi Ueda 已提交
2228 2229
	if (md->bs)
		bioset_free(md->bs);
2230

M
Mikulas Patocka 已提交
2231
	cleanup_srcu_struct(&md->io_barrier);
2232
	free_table_devices(&md->table_devices);
2233
	dm_stats_cleanup(&md->stats);
J
Jeff Mahoney 已提交
2234 2235 2236 2237

	spin_lock(&_minor_lock);
	md->disk->private_data = NULL;
	spin_unlock(&_minor_lock);
2238 2239 2240
	if (blk_get_integrity(md->disk))
		blk_integrity_unregister(md->disk);
	del_gendisk(md->disk);
L
Linus Torvalds 已提交
2241
	put_disk(md->disk);
2242
	blk_cleanup_queue(md->queue);
2243 2244 2245
	bdput(md->bdev);
	free_minor(minor);

2246
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
2247 2248 2249
	kfree(md);
}

K
Kiyoshi Ueda 已提交
2250 2251
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
2252
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
2253

J
Jun'ichi Nomura 已提交
2254
	if (md->io_pool && md->bs) {
2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
		/* The md already has necessary mempools. */
		if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) {
			/*
			 * Reload bioset because front_pad may have changed
			 * because a different table was loaded.
			 */
			bioset_free(md->bs);
			md->bs = p->bs;
			p->bs = NULL;
		}
2265 2266 2267 2268 2269 2270 2271 2272
		/*
		 * There's no need to reload with request-based dm
		 * because the size of front_pad doesn't change.
		 * Note for future: If you are to reload bioset,
		 * prep-ed requests in the queue may refer
		 * to bio from the old bioset, so you must walk
		 * through the queue to unprep.
		 */
K
Kiyoshi Ueda 已提交
2273
		goto out;
M
Mikulas Patocka 已提交
2274
	}
K
Kiyoshi Ueda 已提交
2275

2276
	BUG_ON(!p || md->io_pool || md->rq_pool || md->bs);
K
Kiyoshi Ueda 已提交
2277 2278 2279

	md->io_pool = p->io_pool;
	p->io_pool = NULL;
2280 2281
	md->rq_pool = p->rq_pool;
	p->rq_pool = NULL;
K
Kiyoshi Ueda 已提交
2282 2283 2284 2285 2286 2287 2288 2289
	md->bs = p->bs;
	p->bs = NULL;

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

L
Linus Torvalds 已提交
2290 2291 2292 2293 2294
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2295 2296
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2297 2298
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2299 2300 2301 2302
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2303
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2304

L
Linus Torvalds 已提交
2305 2306 2307 2308
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2309 2310 2311
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2312
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2313
{
2314
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2315

2316
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2317 2318
}

2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
/*
 * Return 1 if the queue has a compulsory merge_bvec_fn function.
 *
 * If this function returns 0, then the device is either a non-dm
 * device without a merge_bvec_fn, or it is a dm device that is
 * able to split any bios it receives that are too big.
 */
int dm_queue_merge_is_compulsory(struct request_queue *q)
{
	struct mapped_device *dev_md;

	if (!q->merge_bvec_fn)
		return 0;

2333
	if (q->make_request_fn == dm_make_request) {
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 2366 2367 2368 2369 2370 2371
		dev_md = q->queuedata;
		if (test_bit(DMF_MERGE_IS_OPTIONAL, &dev_md->flags))
			return 0;
	}

	return 1;
}

static int dm_device_merge_is_compulsory(struct dm_target *ti,
					 struct dm_dev *dev, sector_t start,
					 sector_t len, void *data)
{
	struct block_device *bdev = dev->bdev;
	struct request_queue *q = bdev_get_queue(bdev);

	return dm_queue_merge_is_compulsory(q);
}

/*
 * Return 1 if it is acceptable to ignore merge_bvec_fn based
 * on the properties of the underlying devices.
 */
static int dm_table_merge_is_optional(struct dm_table *table)
{
	unsigned i = 0;
	struct dm_target *ti;

	while (i < dm_table_get_num_targets(table)) {
		ti = dm_table_get_target(table, i++);

		if (ti->type->iterate_devices &&
		    ti->type->iterate_devices(ti, dm_device_merge_is_compulsory, NULL))
			return 0;
	}

	return 1;
}

2372 2373 2374 2375 2376
/*
 * 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
2377
{
2378
	struct dm_table *old_map;
2379
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2380
	sector_t size;
2381
	int merge_is_optional;
L
Linus Torvalds 已提交
2382 2383

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2384 2385 2386 2387

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
2388
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
2389 2390
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
2391
	__set_size(md, size);
2392

2393 2394
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406
	/*
	 * The queue hasn't been stopped yet, if the old table type wasn't
	 * for request-based during suspension.  So stop it to prevent
	 * I/O mapping before resume.
	 * This must be done before setting the queue restrictions,
	 * because request-based dm may be run just after the setting.
	 */
	if (dm_table_request_based(t) && !blk_queue_stopped(q))
		stop_queue(q);

	__bind_mempools(md, t);

2407 2408
	merge_is_optional = dm_table_merge_is_optional(t);

2409
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
M
Mikulas Patocka 已提交
2410
	rcu_assign_pointer(md->map, t);
2411 2412
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2413
	dm_table_set_restrictions(t, q, limits);
2414 2415 2416 2417
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
2418 2419
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
2420

2421
	return old_map;
L
Linus Torvalds 已提交
2422 2423
}

2424 2425 2426 2427
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2428
{
2429
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
2430 2431

	if (!map)
2432
		return NULL;
L
Linus Torvalds 已提交
2433 2434

	dm_table_event_callback(map, NULL, NULL);
2435
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
2436
	dm_sync_table(md);
2437 2438

	return map;
L
Linus Torvalds 已提交
2439 2440 2441 2442 2443
}

/*
 * Constructor for a new device.
 */
2444
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2445 2446 2447
{
	struct mapped_device *md;

2448
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2449 2450 2451
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2452 2453
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2454 2455 2456 2457
	*result = md;
	return 0;
}

2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
/*
 * 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)
{
2474
	BUG_ON(!mutex_is_locked(&md->type_lock));
2475 2476 2477 2478 2479
	md->type = type;
}

unsigned dm_get_md_type(struct mapped_device *md)
{
2480
	BUG_ON(!mutex_is_locked(&md->type_lock));
2481 2482 2483
	return md->type;
}

2484 2485 2486 2487 2488 2489 2490 2491
static bool dm_md_type_request_based(struct mapped_device *md)
{
	unsigned table_type = dm_get_md_type(md);

	return (table_type == DM_TYPE_REQUEST_BASED ||
		table_type == DM_TYPE_MQ_REQUEST_BASED);
}

2492 2493 2494 2495 2496
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507
/*
 * 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);

2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
/*
 * Fully initialize a request-based queue (->elevator, ->request_fn, etc).
 */
static int dm_init_request_based_queue(struct mapped_device *md)
{
	struct request_queue *q = NULL;

	if (md->queue->elevator)
		return 1;

	/* Fully initialize the queue */
	q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL);
	if (!q)
		return 0;

	md->queue = q;
	dm_init_md_queue(md);
	blk_queue_softirq_done(md->queue, dm_softirq_done);
	blk_queue_prep_rq(md->queue, dm_prep_fn);

2528 2529 2530 2531 2532
	/* Also initialize the request-based DM worker thread */
	init_kthread_worker(&md->kworker);
	md->kworker_task = kthread_run(kthread_worker_fn, &md->kworker,
				       "kdmwork-%s", dm_device_name(md));

2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
	elv_register_queue(md->queue);

	return 1;
}

/*
 * Setup the DM device's queue based on md's type
 */
int dm_setup_md_queue(struct mapped_device *md)
{
2543 2544 2545 2546 2547 2548 2549 2550 2551
	if (dm_md_type_request_based(md)) {
		if (!dm_init_request_based_queue(md)) {
			DMWARN("Cannot initialize queue for request-based mapped device");
			return -EINVAL;
		}
	} else {
		/* bio-based specific initialization */
		blk_queue_make_request(md->queue, dm_make_request);
		blk_queue_merge_bvec(md->queue, dm_merge_bvec);
2552 2553 2554 2555 2556
	}

	return 0;
}

2557
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
2558 2559 2560 2561 2562 2563 2564
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2565
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2566 2567

	md = idr_find(&_minor_idr, minor);
2568 2569 2570 2571 2572 2573 2574 2575 2576
	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
2577
	}
L
Linus Torvalds 已提交
2578

J
Jeff Mahoney 已提交
2579
out:
2580
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2581

2582 2583
	return md;
}
A
Alasdair G Kergon 已提交
2584
EXPORT_SYMBOL_GPL(dm_get_md);
2585

A
Alasdair G Kergon 已提交
2586
void *dm_get_mdptr(struct mapped_device *md)
2587
{
A
Alasdair G Kergon 已提交
2588
	return md->interface_ptr;
L
Linus Torvalds 已提交
2589 2590 2591 2592 2593 2594 2595 2596 2597 2598
}

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);
2599
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2600 2601
}

2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614
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);

2615 2616 2617 2618 2619 2620
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2621
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2622
{
M
Mike Anderson 已提交
2623
	struct dm_table *map;
M
Mikulas Patocka 已提交
2624
	int srcu_idx;
L
Linus Torvalds 已提交
2625

2626
	might_sleep();
J
Jeff Mahoney 已提交
2627

M
Mikulas Patocka 已提交
2628
	map = dm_get_live_table(md, &srcu_idx);
2629 2630

	spin_lock(&_minor_lock);
2631 2632 2633 2634
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);

2635 2636 2637
	if (dm_request_based(md))
		flush_kthread_worker(&md->kworker);

2638 2639 2640 2641 2642
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
2643 2644 2645
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2646
	}
2647
	mutex_unlock(&md->suspend_lock);
2648

M
Mikulas Patocka 已提交
2649 2650 2651
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);

2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682
	/*
	 * 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
2683
}
E
Edward Goggin 已提交
2684
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2685

2686
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2687 2688
{
	int r = 0;
2689 2690 2691
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2692 2693

	while (1) {
2694
		set_current_state(interruptible);
2695

2696
		if (!md_in_flight(md))
2697 2698
			break;

2699 2700
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2701 2702 2703 2704 2705 2706 2707 2708
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2709 2710
	remove_wait_queue(&md->wait, &wait);

2711 2712 2713
	return r;
}

L
Linus Torvalds 已提交
2714 2715 2716
/*
 * Process the deferred bios
 */
2717
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2718
{
2719 2720
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2721
	struct bio *c;
M
Mikulas Patocka 已提交
2722 2723
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2724

M
Mikulas Patocka 已提交
2725
	map = dm_get_live_table(md, &srcu_idx);
2726

2727
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2728 2729 2730 2731
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2732
		if (!c)
A
Alasdair G Kergon 已提交
2733
			break;
2734

K
Kiyoshi Ueda 已提交
2735 2736
		if (dm_request_based(md))
			generic_make_request(c);
2737
		else
M
Mikulas Patocka 已提交
2738
			__split_and_process_bio(md, map, c);
2739
	}
M
Milan Broz 已提交
2740

M
Mikulas Patocka 已提交
2741
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2742 2743
}

2744
static void dm_queue_flush(struct mapped_device *md)
2745
{
2746
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2747
	smp_mb__after_atomic();
2748
	queue_work(md->wq, &md->work);
2749 2750
}

L
Linus Torvalds 已提交
2751
/*
2752
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2753
 */
2754
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2755
{
2756
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2757
	struct queue_limits limits;
2758
	int r;
L
Linus Torvalds 已提交
2759

2760
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2761 2762

	/* device must be suspended */
2763
	if (!dm_suspended_md(md))
2764
		goto out;
L
Linus Torvalds 已提交
2765

2766 2767 2768 2769 2770 2771 2772
	/*
	 * 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
2773
		live_map = dm_get_live_table_fast(md);
2774 2775
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2776
		dm_put_live_table_fast(md);
2777 2778
	}

2779 2780 2781 2782 2783 2784
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2785
	}
2786

2787
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2788

2789
out:
2790
	mutex_unlock(&md->suspend_lock);
2791
	return map;
L
Linus Torvalds 已提交
2792 2793 2794 2795 2796 2797
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2798
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2799
{
2800
	int r;
L
Linus Torvalds 已提交
2801 2802

	WARN_ON(md->frozen_sb);
2803

2804
	md->frozen_sb = freeze_bdev(md->bdev);
2805
	if (IS_ERR(md->frozen_sb)) {
2806
		r = PTR_ERR(md->frozen_sb);
2807 2808
		md->frozen_sb = NULL;
		return r;
2809 2810
	}

2811 2812
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2813 2814 2815
	return 0;
}

2816
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2817
{
2818 2819 2820
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2821
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2822
	md->frozen_sb = NULL;
2823
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2824 2825 2826
}

/*
2827 2828 2829
 * 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.
2830
 *
2831
 * Caller must hold md->suspend_lock
2832
 */
2833 2834
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
			unsigned suspend_flags, int interruptible)
L
Linus Torvalds 已提交
2835
{
2836 2837 2838
	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
2839

2840 2841 2842 2843 2844 2845 2846
	/*
	 * 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);

2847 2848 2849 2850
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
2851 2852
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2853
	/*
K
Kiyoshi Ueda 已提交
2854 2855 2856 2857
	 * 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
2858 2859 2860
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
2861 2862
		if (r) {
			dm_table_presuspend_undo_targets(map);
2863
			return r;
2864
		}
2865
	}
L
Linus Torvalds 已提交
2866 2867

	/*
2868 2869 2870 2871 2872 2873 2874
	 * 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
2875 2876 2877
	 * __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
2878
	 */
2879
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2880 2881
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2882

2883
	/*
2884 2885
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2886
	 */
2887
	if (dm_request_based(md)) {
K
Kiyoshi Ueda 已提交
2888
		stop_queue(md->queue);
2889 2890
		flush_kthread_worker(&md->kworker);
	}
2891

2892 2893
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2894
	/*
2895 2896 2897
	 * 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
2898
	 */
2899
	r = dm_wait_for_completion(md, interruptible);
L
Linus Torvalds 已提交
2900

2901
	if (noflush)
2902
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2903 2904
	if (map)
		synchronize_srcu(&md->io_barrier);
2905

L
Linus Torvalds 已提交
2906
	/* were we interrupted ? */
2907
	if (r < 0) {
2908
		dm_queue_flush(md);
M
Milan Broz 已提交
2909

2910
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2911
			start_queue(md->queue);
2912

2913
		unlock_fs(md);
2914
		dm_table_presuspend_undo_targets(map);
2915
		/* pushback list is already flushed, so skip flush */
2916
	}
L
Linus Torvalds 已提交
2917

2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958
	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;
	}

2959
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
2960 2961 2962 2963

	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE);
	if (r)
		goto out_unlock;
2964

2965
	set_bit(DMF_SUSPENDED, &md->flags);
2966

2967 2968
	dm_table_postsuspend_targets(map);

2969
out_unlock:
2970
	mutex_unlock(&md->suspend_lock);
2971
	return r;
L
Linus Torvalds 已提交
2972 2973
}

2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996
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))
		start_queue(md->queue);

	unlock_fs(md);

	return 0;
}

L
Linus Torvalds 已提交
2997 2998
int dm_resume(struct mapped_device *md)
{
2999 3000
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
3001

3002 3003 3004
retry:
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

3005
	if (!dm_suspended_md(md))
3006 3007
		goto out;

3008 3009 3010 3011 3012 3013 3014 3015 3016
	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;
	}

3017
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
3018
	if (!map || !dm_table_get_size(map))
3019
		goto out;
L
Linus Torvalds 已提交
3020

3021
	r = __dm_resume(md, map);
3022 3023
	if (r)
		goto out;
3024 3025 3026

	clear_bit(DMF_SUSPENDED, &md->flags);

3027 3028
	r = 0;
out:
3029
	mutex_unlock(&md->suspend_lock);
3030

3031
	return r;
L
Linus Torvalds 已提交
3032 3033
}

M
Mikulas Patocka 已提交
3034 3035 3036 3037 3038 3039
/*
 * 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.
 */

3040
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
3041
{
3042 3043
	struct dm_table *map = NULL;

3044
	if (md->internal_suspend_count++)
3045 3046 3047 3048 3049 3050 3051
		return; /* nested internal suspend */

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

3052
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068

	/*
	 * 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.
	 */
	(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE);

	set_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);

	dm_table_postsuspend_targets(map);
}

static void __dm_internal_resume(struct mapped_device *md)
{
3069 3070 3071
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
3072 3073
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
3074
	if (dm_suspended_md(md))
3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113
		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
3114 3115 3116 3117 3118 3119 3120
		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);
}
3121
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
3122

3123
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
3124
{
3125
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
3126 3127 3128 3129 3130 3131 3132
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
3133
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
3134

L
Linus Torvalds 已提交
3135 3136 3137
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
3138
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
3139
		       unsigned cookie)
3140
{
M
Milan Broz 已提交
3141 3142 3143 3144
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
3145
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
3146 3147 3148
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
3149 3150
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
3151
	}
3152 3153
}

M
Mike Anderson 已提交
3154 3155 3156 3157 3158
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169
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
3170 3171 3172 3173 3174 3175 3176 3177 3178
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
3179 3180 3181 3182 3183 3184 3185 3186 3187
/*
 * The gendisk is only valid as long as you have a reference
 * count on 'md'.
 */
struct gendisk *dm_disk(struct mapped_device *md)
{
	return md->disk;
}

M
Milan Broz 已提交
3188 3189
struct kobject *dm_kobject(struct mapped_device *md)
{
3190
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
3191 3192 3193 3194 3195 3196
}

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

3197
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
3198

3199
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
3200
	    dm_deleting_md(md))
3201 3202
		return NULL;

M
Milan Broz 已提交
3203 3204 3205 3206
	dm_get(md);
	return md;
}

3207
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
3208 3209 3210 3211
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

3212 3213 3214 3215 3216
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
3217 3218 3219 3220 3221
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

3222 3223
int dm_suspended(struct dm_target *ti)
{
3224
	return dm_suspended_md(dm_table_get_md(ti->table));
3225 3226 3227
}
EXPORT_SYMBOL_GPL(dm_suspended);

3228 3229
int dm_noflush_suspending(struct dm_target *ti)
{
3230
	return __noflush_suspending(dm_table_get_md(ti->table));
3231 3232 3233
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

M
Mikulas Patocka 已提交
3234
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size)
K
Kiyoshi Ueda 已提交
3235
{
J
Jun'ichi Nomura 已提交
3236 3237
	struct dm_md_mempools *pools = kzalloc(sizeof(*pools), GFP_KERNEL);
	struct kmem_cache *cachep;
3238
	unsigned int pool_size = 0;
J
Jun'ichi Nomura 已提交
3239
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
3240 3241 3242 3243

	if (!pools)
		return NULL;

3244 3245
	switch (type) {
	case DM_TYPE_BIO_BASED:
J
Jun'ichi Nomura 已提交
3246
		cachep = _io_cache;
3247
		pool_size = dm_get_reserved_bio_based_ios();
J
Jun'ichi Nomura 已提交
3248
		front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
3249 3250
		break;
	case DM_TYPE_REQUEST_BASED:
3251
		pool_size = dm_get_reserved_rq_based_ios();
3252 3253 3254
		pools->rq_pool = mempool_create_slab_pool(pool_size, _rq_cache);
		if (!pools->rq_pool)
			goto out;
3255 3256 3257 3258 3259
		/* fall through to setup remaining rq-based pools */
	case DM_TYPE_MQ_REQUEST_BASED:
		cachep = _rq_tio_cache;
		if (!pool_size)
			pool_size = dm_get_reserved_rq_based_ios();
J
Jun'ichi Nomura 已提交
3260 3261 3262
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
		/* per_bio_data_size is not used. See __bind_mempools(). */
		WARN_ON(per_bio_data_size != 0);
3263 3264
		break;
	default:
J
Jun'ichi Nomura 已提交
3265
		goto out;
3266
	}
K
Kiyoshi Ueda 已提交
3267

3268
	pools->io_pool = mempool_create_slab_pool(pool_size, cachep);
J
Jun'ichi Nomura 已提交
3269 3270
	if (!pools->io_pool)
		goto out;
K
Kiyoshi Ueda 已提交
3271

J
Junichi Nomura 已提交
3272
	pools->bs = bioset_create_nobvec(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
3273
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
3274
		goto out;
K
Kiyoshi Ueda 已提交
3275

3276
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
3277
		goto out;
3278

K
Kiyoshi Ueda 已提交
3279 3280
	return pools;

J
Jun'ichi Nomura 已提交
3281 3282
out:
	dm_free_md_mempools(pools);
K
Kiyoshi Ueda 已提交
3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294

	return NULL;
}

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

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

3295 3296 3297
	if (pools->rq_pool)
		mempool_destroy(pools->rq_pool);

K
Kiyoshi Ueda 已提交
3298 3299 3300 3301 3302 3303
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

3304
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
3305 3306
	.open = dm_blk_open,
	.release = dm_blk_close,
3307
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
3308
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319
	.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");
3320

3321 3322 3323
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

3324 3325 3326
module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");

L
Linus Torvalds 已提交
3327 3328 3329
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");