dm.c 82.2 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
#include <linux/ktime.h>
25
#include <linux/elevator.h> /* for rq_end_sector() */
26
#include <linux/blk-mq.h>
27 28

#include <trace/events/block.h>
L
Linus Torvalds 已提交
29

30 31
#define DM_MSG_PREFIX "core"

N
Namhyung Kim 已提交
32 33 34 35 36 37 38 39 40 41
#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
42 43 44 45 46 47 48
/*
 * 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
49 50 51 52 53
static const char *_name = DM_NAME;

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

54 55
static DEFINE_IDR(_minor_idr);

56
static DEFINE_SPINLOCK(_minor_lock);
M
Mikulas Patocka 已提交
57 58 59 60 61

static void do_deferred_remove(struct work_struct *w);

static DECLARE_WORK(deferred_remove_work, do_deferred_remove);

62 63
static struct workqueue_struct *deferred_remove_workqueue;

L
Linus Torvalds 已提交
64
/*
K
Kiyoshi Ueda 已提交
65
 * For bio-based dm.
L
Linus Torvalds 已提交
66 67 68 69 70 71
 * 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
72
	struct bio *bio;
73
	unsigned long start_time;
74
	spinlock_t endio_lock;
M
Mikulas Patocka 已提交
75
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
76 77
};

K
Kiyoshi Ueda 已提交
78 79 80 81 82 83 84
/*
 * For request-based dm.
 * One of these is allocated per request.
 */
struct dm_rq_target_io {
	struct mapped_device *md;
	struct dm_target *ti;
85
	struct request *orig, *clone;
86
	struct kthread_work work;
K
Kiyoshi Ueda 已提交
87 88 89 90 91
	int error;
	union map_info info;
};

/*
92 93 94 95 96 97
 * 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
98 99 100
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
101
	struct dm_rq_target_io *tio;
102
	struct bio clone;
K
Kiyoshi Ueda 已提交
103 104
};

105 106 107 108 109 110 111 112
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);

113 114
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
115 116 117
/*
 * Bits for the md->flags field.
 */
118
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
119
#define DMF_SUSPENDED 1
120
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
121
#define DMF_FREEING 3
122
#define DMF_DELETING 4
123
#define DMF_NOFLUSH_SUSPENDING 5
124
#define DMF_MERGE_IS_OPTIONAL 6
M
Mikulas Patocka 已提交
125
#define DMF_DEFERRED_REMOVE 7
126
#define DMF_SUSPENDED_INTERNALLY 8
L
Linus Torvalds 已提交
127

M
Mikulas Patocka 已提交
128 129 130 131 132 133 134 135
/*
 * A dummy definition to make RCU happy.
 * struct dm_table should never be dereferenced in this file.
 */
struct dm_table {
	int undefined__;
};

136 137 138
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
139
struct mapped_device {
M
Mikulas Patocka 已提交
140
	struct srcu_struct io_barrier;
141
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
142
	atomic_t holders;
143
	atomic_t open_count;
L
Linus Torvalds 已提交
144

M
Mikulas Patocka 已提交
145 146 147 148 149
	/*
	 * The current mapping.
	 * Use dm_get_live_table{_fast} or take suspend_lock for
	 * dereference.
	 */
150
	struct dm_table __rcu *map;
M
Mikulas Patocka 已提交
151

152 153 154
	struct list_head table_devices;
	struct mutex table_devices_lock;

L
Linus Torvalds 已提交
155 156
	unsigned long flags;

157
	struct request_queue *queue;
158
	unsigned type;
159
	/* Protect queue and type against concurrent access. */
160 161
	struct mutex type_lock;

162 163
	struct target_type *immutable_target_type;

L
Linus Torvalds 已提交
164
	struct gendisk *disk;
M
Mike Anderson 已提交
165
	char name[16];
L
Linus Torvalds 已提交
166 167 168 169 170 171

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
172
	atomic_t pending[2];
L
Linus Torvalds 已提交
173
	wait_queue_head_t wait;
174
	struct work_struct work;
K
Kiyoshi Ueda 已提交
175
	struct bio_list deferred;
176
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
177

178
	/*
179
	 * Processing queue (flush)
180 181 182
	 */
	struct workqueue_struct *wq;

L
Linus Torvalds 已提交
183 184 185 186
	/*
	 * io objects are allocated from here.
	 */
	mempool_t *io_pool;
187
	mempool_t *rq_pool;
L
Linus Torvalds 已提交
188

S
Stefan Bader 已提交
189 190
	struct bio_set *bs;

L
Linus Torvalds 已提交
191 192 193 194 195
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
196 197 198
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
199 200 201 202 203

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
204
	struct block_device *bdev;
D
Darrick J. Wong 已提交
205 206 207

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
208

209 210
	/* kobject and completion */
	struct dm_kobject_holder kobj_holder;
211

212 213
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
M
Mikulas Patocka 已提交
214

215 216 217
	/* the number of internal suspends */
	unsigned internal_suspend_count;

M
Mikulas Patocka 已提交
218
	struct dm_stats stats;
219 220 221

	struct kthread_worker kworker;
	struct task_struct *kworker_task;
222 223

	/* for request-based merge heuristic in dm_request_fn() */
224
	unsigned seq_rq_merge_deadline_usecs;
225
	int last_rq_rw;
226 227
	sector_t last_rq_pos;
	ktime_t last_rq_start_time;
228 229 230

	/* for blk-mq request-based DM support */
	struct blk_mq_tag_set tag_set;
231
	bool use_blk_mq;
L
Linus Torvalds 已提交
232 233
};

234 235 236 237 238 239 240 241 242 243 244
#ifdef CONFIG_DM_MQ_DEFAULT
static bool use_blk_mq = true;
#else
static bool use_blk_mq = false;
#endif

bool dm_use_blk_mq(struct mapped_device *md)
{
	return md->use_blk_mq;
}

K
Kiyoshi Ueda 已提交
245 246 247 248 249
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
250
	mempool_t *rq_pool;
K
Kiyoshi Ueda 已提交
251 252 253
	struct bio_set *bs;
};

254 255 256 257 258 259
struct table_device {
	struct list_head list;
	atomic_t count;
	struct dm_dev dm_dev;
};

260 261
#define RESERVED_BIO_BASED_IOS		16
#define RESERVED_REQUEST_BASED_IOS	256
262
#define RESERVED_MAX_IOS		1024
263
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
264
static struct kmem_cache *_rq_tio_cache;
265
static struct kmem_cache *_rq_cache;
266

267 268 269 270 271
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

272 273 274 275 276
/*
 * Request-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;

277
static unsigned __dm_get_module_param(unsigned *module_param,
278 279
				      unsigned def, unsigned max)
{
280 281
	unsigned param = ACCESS_ONCE(*module_param);
	unsigned modified_param = 0;
282

283 284 285 286
	if (!param)
		modified_param = def;
	else if (param > max)
		modified_param = max;
287

288 289 290
	if (modified_param) {
		(void)cmpxchg(module_param, param, modified_param);
		param = modified_param;
291 292
	}

293
	return param;
294 295
}

296 297
unsigned dm_get_reserved_bio_based_ios(void)
{
298
	return __dm_get_module_param(&reserved_bio_based_ios,
299 300 301 302
				     RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

303 304
unsigned dm_get_reserved_rq_based_ios(void)
{
305
	return __dm_get_module_param(&reserved_rq_based_ios,
306 307 308 309
				     RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);

L
Linus Torvalds 已提交
310 311
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
312
	int r = -ENOMEM;
L
Linus Torvalds 已提交
313 314

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
315
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
316
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
317
		return r;
L
Linus Torvalds 已提交
318

K
Kiyoshi Ueda 已提交
319 320
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
321
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
322

323 324 325 326 327
	_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
328
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
329
	if (r)
330
		goto out_free_rq_cache;
M
Mike Anderson 已提交
331

332 333 334 335 336 337
	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
338 339
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
340
	if (r < 0)
341
		goto out_free_workqueue;
L
Linus Torvalds 已提交
342 343 344 345 346

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
347

348 349
out_free_workqueue:
	destroy_workqueue(deferred_remove_workqueue);
K
Kiyoshi Ueda 已提交
350 351
out_uevent_exit:
	dm_uevent_exit();
352 353
out_free_rq_cache:
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
354 355
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
356 357 358 359
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
360 361 362 363
}

static void local_exit(void)
{
M
Mikulas Patocka 已提交
364
	flush_scheduled_work();
365
	destroy_workqueue(deferred_remove_workqueue);
M
Mikulas Patocka 已提交
366

367
	kmem_cache_destroy(_rq_cache);
K
Kiyoshi Ueda 已提交
368
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
369
	kmem_cache_destroy(_io_cache);
370
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
371
	dm_uevent_exit();
L
Linus Torvalds 已提交
372 373 374 375 376 377

	_major = 0;

	DMINFO("cleaned up");
}

378
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
379 380 381 382
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
383
	dm_io_init,
384
	dm_kcopyd_init,
L
Linus Torvalds 已提交
385
	dm_interface_init,
M
Mikulas Patocka 已提交
386
	dm_statistics_init,
L
Linus Torvalds 已提交
387 388
};

389
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
390 391 392 393
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
394
	dm_io_exit,
395
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
396
	dm_interface_exit,
M
Mikulas Patocka 已提交
397
	dm_statistics_exit,
L
Linus Torvalds 已提交
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
};

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]();
427 428 429 430 431

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
432 433 434 435 436
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
437 438 439 440 441
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
442
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
443 444 445
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
446 447
	spin_lock(&_minor_lock);

A
Al Viro 已提交
448
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
449 450 451
	if (!md)
		goto out;

452
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
453
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
454 455 456 457
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
458
	dm_get(md);
459
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
460 461 462 463
out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
464 465
}

466
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
467
{
468
	struct mapped_device *md;
469

470 471
	spin_lock(&_minor_lock);

472 473 474 475
	md = disk->private_data;
	if (WARN_ON(!md))
		goto out;

M
Mikulas Patocka 已提交
476 477
	if (atomic_dec_and_test(&md->open_count) &&
	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
478
		queue_work(deferred_remove_workqueue, &deferred_remove_work);
M
Mikulas Patocka 已提交
479

L
Linus Torvalds 已提交
480
	dm_put(md);
481
out:
482
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
483 484
}

485 486 487 488 489 490 491 492
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
493
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
494 495 496 497 498
{
	int r = 0;

	spin_lock(&_minor_lock);

M
Mikulas Patocka 已提交
499
	if (dm_open_count(md)) {
500
		r = -EBUSY;
M
Mikulas Patocka 已提交
501 502 503 504
		if (mark_deferred)
			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
		r = -EEXIST;
505 506 507 508 509 510 511 512
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

M
Mikulas Patocka 已提交
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
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
534 535 536 537 538
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

539 540 541 542 543
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
	return md->queue;
}

M
Mikulas Patocka 已提交
544 545 546 547 548
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
549 550 551 552 553 554 555
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
556
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
557 558
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
559
	struct mapped_device *md = bdev->bd_disk->private_data;
M
Mikulas Patocka 已提交
560
	int srcu_idx;
561
	struct dm_table *map;
562 563 564
	struct dm_target *tgt;
	int r = -ENOTTY;

565
retry:
M
Mikulas Patocka 已提交
566 567
	map = dm_get_live_table(md, &srcu_idx);

568 569 570 571 572 573 574 575
	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);
576 577
	if (!tgt->type->ioctl)
		goto out;
578

579
	if (dm_suspended_md(md)) {
580 581 582 583
		r = -EAGAIN;
		goto out;
	}

584
	r = tgt->type->ioctl(tgt, cmd, arg);
585 586

out:
M
Mikulas Patocka 已提交
587
	dm_put_live_table(md, srcu_idx);
588

589 590 591 592 593
	if (r == -ENOTCONN) {
		msleep(10);
		goto retry;
	}

594 595 596
	return r;
}

A
Alasdair G Kergon 已提交
597
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
598 599 600 601
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
602
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
603 604 605 606
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
607
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
608
{
609
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
610 611
}

K
Kiyoshi Ueda 已提交
612 613
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
614
{
J
Jun'ichi Nomura 已提交
615
	return mempool_alloc(md->io_pool, gfp_mask);
616 617 618 619
}

static void free_rq_tio(struct dm_rq_target_io *tio)
{
J
Jun'ichi Nomura 已提交
620
	mempool_free(tio, tio->md->io_pool);
621 622
}

623 624 625 626 627 628 629 630 631 632 633
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
634 635 636 637 638 639
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

640 641 642
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
643
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
644
	int cpu;
M
Mikulas Patocka 已提交
645
	int rw = bio_data_dir(bio);
646 647 648

	io->start_time = jiffies;

T
Tejun Heo 已提交
649 650 651
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
652 653
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
654 655

	if (unlikely(dm_stats_used(&md->stats)))
656
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
M
Mikulas Patocka 已提交
657
				    bio_sectors(bio), false, 0, &io->stats_aux);
658 659
}

660
static void end_io_acct(struct dm_io *io)
661 662 663 664
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
665
	int pending;
666 667
	int rw = bio_data_dir(bio);

668
	generic_end_io_acct(rw, &dm_disk(md)->part0, io->start_time);
669

M
Mikulas Patocka 已提交
670
	if (unlikely(dm_stats_used(&md->stats)))
671
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
M
Mikulas Patocka 已提交
672 673
				    bio_sectors(bio), true, duration, &io->stats_aux);

674 675
	/*
	 * After this is decremented the bio must not be touched if it is
676
	 * a flush.
677
	 */
678 679
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
680
	pending += atomic_read(&md->pending[rw^0x1]);
681

682 683 684
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
685 686
}

L
Linus Torvalds 已提交
687 688 689
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
690
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
691
{
692
	unsigned long flags;
L
Linus Torvalds 已提交
693

694
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
695
	bio_list_add(&md->deferred, bio);
696
	spin_unlock_irqrestore(&md->deferred_lock, flags);
697
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
698 699 700 701 702
}

/*
 * 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
703
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
704
 */
M
Mikulas Patocka 已提交
705
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
706
{
M
Mikulas Patocka 已提交
707 708 709 710
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
711

M
Mikulas Patocka 已提交
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731
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
732

M
Mikulas Patocka 已提交
733 734 735
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
736 737
}

738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
/*
 * 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
852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
/*
 * 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
879 880 881 882 883 884 885 886 887
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

888 889 890 891 892
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
893 894 895 896
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
897
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
898
{
899
	unsigned long flags;
900 901 902
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
903 904

	/* Push-back supersedes any I/O errors */
905 906 907 908 909 910
	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
911 912

	if (atomic_dec_and_test(&io->io_count)) {
913 914 915 916
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
917
			spin_lock_irqsave(&md->deferred_lock, flags);
918 919 920
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
921 922
				/* noflush suspend was interrupted. */
				io->error = -EIO;
923
			spin_unlock_irqrestore(&md->deferred_lock, flags);
924 925
		}

926 927
		io_error = io->error;
		bio = io->bio;
928 929 930 931 932
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
933

934
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_iter.bi_size) {
935
			/*
936 937
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
938
			 */
939 940
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
941
		} else {
942
			/* done with normal IO or empty flush */
943
			trace_block_bio_complete(md->queue, bio, io_error);
944
			bio_endio(bio, io_error);
945
		}
L
Linus Torvalds 已提交
946 947 948
	}
}

949 950 951 952 953 954 955 956
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;
}

957
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
958
{
959
	int r = error;
M
Mikulas Patocka 已提交
960
	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
961
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
962
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
963 964 965 966 967 968
	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
969
		r = endio(tio->ti, bio, error);
970 971 972 973 974
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
975
			error = r;
976 977
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
978
			return;
979 980 981 982
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
983 984
	}

985 986 987 988
	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
989
	free_tio(md, tio);
990
	dec_pending(io, error);
L
Linus Torvalds 已提交
991 992
}

993 994 995 996 997
static struct dm_rq_target_io *tio_from_request(struct request *rq)
{
	return (rq->q->mq_ops ? blk_mq_rq_to_pdu(rq) : rq->special);
}

998 999 1000 1001 1002
/*
 * 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.
 */
1003
static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
1004
{
1005 1006
	int nr_requests_pending;

1007
	atomic_dec(&md->pending[rw]);
1008 1009

	/* nudge anyone waiting on suspend queue */
1010 1011
	nr_requests_pending = md_in_flight(md);
	if (!nr_requests_pending)
1012 1013
		wake_up(&md->wait);

1014 1015 1016 1017 1018 1019
	/*
	 * 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.
	 */
1020
	if (run_queue) {
1021 1022 1023 1024
		if (md->queue->mq_ops)
			blk_mq_run_hw_queues(md->queue, true);
		else if (!nr_requests_pending ||
			 (nr_requests_pending >= md->queue->nr_congestion_on))
1025 1026
			blk_run_queue_async(md->queue);
	}
1027 1028 1029 1030 1031 1032 1033

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

1034
static void free_rq_clone(struct request *clone)
1035 1036
{
	struct dm_rq_target_io *tio = clone->end_io_data;
1037
	struct mapped_device *md = tio->md;
1038

1039 1040
	if (md->type == DM_TYPE_MQ_REQUEST_BASED)
		/* stacked on blk-mq queue(s) */
1041
		tio->ti->type->release_clone_rq(clone);
1042 1043
	else if (!md->queue->mq_ops)
		/* request_fn queue stacked on request_fn queue(s) */
1044
		free_clone_request(md, clone);
1045 1046 1047 1048 1049
	/*
	 * NOTE: for the blk-mq queue stacked on request_fn queue(s) case:
	 * no need to call free_clone_request() because we leverage blk-mq by
	 * allocating the clone at the end of the blk-mq pdu (see: clone_rq)
	 */
1050 1051 1052

	if (!md->queue->mq_ops)
		free_rq_tio(tio);
1053 1054
}

K
Kiyoshi Ueda 已提交
1055 1056
/*
 * Complete the clone and the original request.
1057 1058
 * Must be called without clone's queue lock held,
 * see end_clone_request() for more details.
K
Kiyoshi Ueda 已提交
1059 1060 1061 1062 1063 1064 1065 1066
 */
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;

1067
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
		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;
	}

1080
	free_rq_clone(clone);
1081 1082 1083 1084
	if (!rq->q->mq_ops)
		blk_end_request_all(rq, error);
	else
		blk_mq_end_request(rq, error);
1085
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
1086 1087
}

1088 1089
static void dm_unprep_request(struct request *rq)
{
1090
	struct dm_rq_target_io *tio = tio_from_request(rq);
1091
	struct request *clone = tio->clone;
1092

1093 1094 1095 1096
	if (!rq->q->mq_ops) {
		rq->special = NULL;
		rq->cmd_flags &= ~REQ_DONTPREP;
	}
1097

1098
	if (clone)
1099
		free_rq_clone(clone);
1100 1101 1102 1103 1104
}

/*
 * Requeue the original request of a clone.
 */
1105
static void old_requeue_request(struct request *rq)
1106 1107 1108 1109 1110 1111
{
	struct request_queue *q = rq->q;
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	blk_requeue_request(q, rq);
J
Junichi Nomura 已提交
1112
	blk_run_queue_async(q);
1113
	spin_unlock_irqrestore(q->queue_lock, flags);
1114 1115
}

1116 1117
static void dm_requeue_original_request(struct mapped_device *md,
					struct request *rq)
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
{
	int rw = rq_data_dir(rq);

	dm_unprep_request(rq);

	if (!rq->q->mq_ops)
		old_requeue_request(rq);
	else {
		blk_mq_requeue_request(rq);
		blk_mq_kick_requeue_list(rq->q);
	}
1129

1130 1131 1132
	rq_completed(md, rw, false);
}

1133
static void old_stop_queue(struct request_queue *q)
1134 1135 1136
{
	unsigned long flags;

1137 1138 1139
	if (blk_queue_stopped(q))
		return;

1140
	spin_lock_irqsave(q->queue_lock, flags);
1141
	blk_stop_queue(q);
1142 1143 1144
	spin_unlock_irqrestore(q->queue_lock, flags);
}

1145
static void stop_queue(struct request_queue *q)
1146
{
1147 1148 1149 1150
	if (!q->mq_ops)
		old_stop_queue(q);
	else
		blk_mq_stop_hw_queues(q);
1151 1152
}

1153
static void old_start_queue(struct request_queue *q)
1154 1155 1156 1157
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
1158 1159
	if (blk_queue_stopped(q))
		blk_start_queue(q);
1160 1161 1162
	spin_unlock_irqrestore(q->queue_lock, flags);
}

1163 1164 1165 1166 1167 1168 1169 1170
static void start_queue(struct request_queue *q)
{
	if (!q->mq_ops)
		old_start_queue(q);
	else
		blk_mq_start_stopped_hw_queues(q, true);
}

1171
static void dm_done(struct request *clone, int error, bool mapped)
1172
{
1173
	int r = error;
1174
	struct dm_rq_target_io *tio = clone->end_io_data;
1175
	dm_request_endio_fn rq_end_io = NULL;
1176

1177 1178 1179 1180 1181 1182
	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);
	}
1183

1184 1185 1186 1187
	if (unlikely(r == -EREMOTEIO && (clone->cmd_flags & REQ_WRITE_SAME) &&
		     !clone->q->limits.max_write_same_sectors))
		disable_write_same(tio->md);

1188
	if (r <= 0)
1189
		/* The target wants to complete the I/O */
1190 1191
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
1192 1193
		/* The target will handle the I/O */
		return;
1194
	else if (r == DM_ENDIO_REQUEUE)
1195
		/* The target wants to requeue the I/O */
1196
		dm_requeue_original_request(tio->md, tio->orig);
1197
	else {
1198
		DMWARN("unimplemented target endio return value: %d", r);
1199 1200 1201 1202
		BUG();
	}
}

1203 1204 1205 1206 1207 1208
/*
 * Request completion handler for request-based dm
 */
static void dm_softirq_done(struct request *rq)
{
	bool mapped = true;
1209
	struct dm_rq_target_io *tio = tio_from_request(rq);
1210
	struct request *clone = tio->clone;
1211
	int rw;
1212

1213
	if (!clone) {
1214 1215 1216 1217 1218 1219 1220 1221 1222
		rw = rq_data_dir(rq);
		if (!rq->q->mq_ops) {
			blk_end_request_all(rq, tio->error);
			rq_completed(tio->md, rw, false);
			free_rq_tio(tio);
		} else {
			blk_mq_end_request(rq, tio->error);
			rq_completed(tio->md, rw, false);
		}
1223 1224
		return;
	}
1225 1226 1227 1228 1229 1230 1231

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

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

1232 1233 1234 1235
/*
 * Complete the clone and the original request with the error status
 * through softirq context.
 */
1236
static void dm_complete_request(struct request *rq, int error)
1237
{
1238
	struct dm_rq_target_io *tio = tio_from_request(rq);
1239 1240 1241 1242 1243 1244 1245 1246 1247

	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.
1248
 * This may be used when the target's map_rq() or clone_and_map_rq() functions fail.
1249
 */
1250
static void dm_kill_unmapped_request(struct request *rq, int error)
1251 1252
{
	rq->cmd_flags |= REQ_FAILED;
1253
	dm_complete_request(rq, error);
1254 1255 1256
}

/*
1257
 * Called with the clone's queue lock held (for non-blk-mq)
1258 1259 1260
 */
static void end_clone_request(struct request *clone, int error)
{
1261 1262
	struct dm_rq_target_io *tio = clone->end_io_data;

1263 1264 1265 1266 1267 1268 1269 1270 1271
	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);
	}
1272 1273 1274

	/*
	 * Actual request completion is done in a softirq context which doesn't
1275
	 * hold the clone's queue lock.  Otherwise, deadlock could occur because:
1276 1277 1278
	 *     - 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
1279
	 *       against this clone's queue
1280
	 */
1281
	dm_complete_request(tio->orig, error);
1282 1283
}

1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
/*
 * 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
1296
{
1297
	sector_t len = max_io_len_target_boundary(sector, ti);
1298
	sector_t offset, max_len;
L
Linus Torvalds 已提交
1299 1300

	/*
1301
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
1302
	 */
1303 1304 1305 1306 1307 1308 1309 1310 1311 1312
	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
1313 1314 1315 1316 1317
	}

	return len;
}

1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
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);

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 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
/*
 * 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
1373
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1374 1375
{
	int r;
1376
	sector_t sector;
S
Stefan Bader 已提交
1377
	struct mapped_device *md;
1378
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1379
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388

	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);
1389
	sector = clone->bi_iter.bi_sector;
M
Mikulas Patocka 已提交
1390
	r = ti->type->map(ti, clone);
1391
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1392
		/* the bio has been remapped so dispatch it */
1393

1394 1395
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
1396

L
Linus Torvalds 已提交
1397
		generic_make_request(clone);
1398 1399
	} 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
1400 1401 1402
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1403 1404 1405
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1406 1407 1408 1409 1410 1411 1412 1413 1414
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
1415
	unsigned sector_count;
L
Linus Torvalds 已提交
1416 1417
};

1418
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
A
Alasdair G Kergon 已提交
1419
{
1420 1421
	bio->bi_iter.bi_sector = sector;
	bio->bi_iter.bi_size = to_bytes(len);
L
Linus Torvalds 已提交
1422 1423 1424 1425 1426
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1427
static void clone_bio(struct dm_target_io *tio, struct bio *bio,
1428
		      sector_t sector, unsigned len)
L
Linus Torvalds 已提交
1429
{
1430
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1431

1432 1433 1434 1435
	__bio_clone_fast(clone, bio);

	if (bio_integrity(bio))
		bio_integrity_clone(clone, bio, GFP_NOIO);
A
Alasdair G Kergon 已提交
1436

1437 1438 1439 1440 1441
	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
1442 1443
}

1444
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1445
				      struct dm_target *ti,
1446
				      unsigned target_bio_nr)
1447
{
1448 1449 1450
	struct dm_target_io *tio;
	struct bio *clone;

1451
	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
1452
	tio = container_of(clone, struct dm_target_io, clone);
1453 1454 1455

	tio->io = ci->io;
	tio->ti = ti;
1456
	tio->target_bio_nr = target_bio_nr;
1457 1458 1459 1460

	return tio;
}

1461 1462
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
1463
				       unsigned target_bio_nr, unsigned *len)
1464
{
1465
	struct dm_target_io *tio = alloc_tio(ci, ti, target_bio_nr);
1466
	struct bio *clone = &tio->clone;
1467

1468 1469
	tio->len_ptr = len;

1470
	__bio_clone_fast(clone, ci->bio);
A
Alasdair G Kergon 已提交
1471
	if (len)
1472
		bio_setup_sector(clone, ci->sector, *len);
1473

A
Alasdair G Kergon 已提交
1474
	__map_bio(tio);
1475 1476
}

1477
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
1478
				  unsigned num_bios, unsigned *len)
1479
{
1480
	unsigned target_bio_nr;
1481

1482
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1483
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1484 1485
}

1486
static int __send_empty_flush(struct clone_info *ci)
1487
{
1488
	unsigned target_nr = 0;
1489 1490
	struct dm_target *ti;

1491
	BUG_ON(bio_has_data(ci->bio));
1492
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1493
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
1494 1495 1496 1497

	return 0;
}

A
Alasdair G Kergon 已提交
1498
static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
1499
				     sector_t sector, unsigned *len)
M
Mike Snitzer 已提交
1500
{
1501
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1502
	struct dm_target_io *tio;
1503 1504
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
M
Mike Snitzer 已提交
1505

1506 1507 1508 1509 1510
	/*
	 * 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
1511

1512
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
1513
		tio = alloc_tio(ci, ti, target_bio_nr);
1514 1515
		tio->len_ptr = len;
		clone_bio(tio, bio, sector, *len);
1516 1517
		__map_bio(tio);
	}
M
Mike Snitzer 已提交
1518 1519
}

1520
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1521

1522
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1523
{
1524
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1525 1526
}

1527
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1528
{
1529
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1530 1531 1532 1533 1534 1535
}

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

static bool is_split_required_for_discard(struct dm_target *ti)
{
1536
	return ti->split_discard_bios;
M
Mike Snitzer 已提交
1537 1538
}

1539 1540 1541
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
1542 1543
{
	struct dm_target *ti;
1544
	unsigned len;
1545
	unsigned num_bios;
M
Mike Snitzer 已提交
1546

1547 1548 1549 1550
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
M
Mike Snitzer 已提交
1551 1552

		/*
M
Mike Snitzer 已提交
1553 1554
		 * 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
1555
		 * reconfiguration might also have changed that since the
1556
		 * check was performed.
M
Mike Snitzer 已提交
1557
		 */
1558 1559
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1560
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1561

M
Mike Snitzer 已提交
1562
		if (is_split_required && !is_split_required(ti))
1563
			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
1564
		else
1565
			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
1566

1567
		__send_duplicate_bios(ci, ti, num_bios, &len);
1568 1569 1570

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1571 1572 1573 1574

	return 0;
}

1575
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1576
{
1577 1578
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1579 1580
}

1581
static int __send_write_same(struct clone_info *ci)
M
Mike Snitzer 已提交
1582
{
1583
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
M
Mike Snitzer 已提交
1584 1585
}

A
Alasdair G Kergon 已提交
1586 1587 1588
/*
 * Select the correct strategy for processing a non-flush bio.
 */
1589
static int __split_and_process_non_flush(struct clone_info *ci)
L
Linus Torvalds 已提交
1590
{
1591
	struct bio *bio = ci->bio;
1592
	struct dm_target *ti;
1593
	unsigned len;
L
Linus Torvalds 已提交
1594

M
Mike Snitzer 已提交
1595
	if (unlikely(bio->bi_rw & REQ_DISCARD))
1596
		return __send_discard(ci);
M
Mike Snitzer 已提交
1597
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
1598
		return __send_write_same(ci);
M
Mike Snitzer 已提交
1599

1600 1601 1602 1603
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1604
	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
L
Linus Torvalds 已提交
1605

1606
	__clone_and_map_data_bio(ci, ti, ci->sector, &len);
L
Linus Torvalds 已提交
1607

1608 1609
	ci->sector += len;
	ci->sector_count -= len;
L
Linus Torvalds 已提交
1610

1611
	return 0;
L
Linus Torvalds 已提交
1612 1613 1614
}

/*
1615
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1616
 */
M
Mikulas Patocka 已提交
1617 1618
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
L
Linus Torvalds 已提交
1619 1620
{
	struct clone_info ci;
1621
	int error = 0;
L
Linus Torvalds 已提交
1622

M
Mikulas Patocka 已提交
1623
	if (unlikely(!map)) {
1624
		bio_io_error(bio);
1625 1626
		return;
	}
1627

M
Mikulas Patocka 已提交
1628
	ci.map = map;
L
Linus Torvalds 已提交
1629 1630 1631 1632 1633 1634
	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;
1635
	spin_lock_init(&ci.io->endio_lock);
1636
	ci.sector = bio->bi_iter.bi_sector;
L
Linus Torvalds 已提交
1637

1638
	start_io_acct(ci.io);
A
Alasdair G Kergon 已提交
1639

1640 1641 1642
	if (bio->bi_rw & REQ_FLUSH) {
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1643
		error = __send_empty_flush(&ci);
1644 1645
		/* dec_pending submits any data associated with flush */
	} else {
1646
		ci.bio = bio;
1647
		ci.sector_count = bio_sectors(bio);
1648
		while (ci.sector_count && !error)
1649
			error = __split_and_process_non_flush(&ci);
1650
	}
L
Linus Torvalds 已提交
1651 1652

	/* drop the extra reference count */
1653
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1654 1655 1656 1657 1658
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1659 1660 1661 1662 1663
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
1664
	struct dm_table *map = dm_get_live_table_fast(md);
M
Milan Broz 已提交
1665
	struct dm_target *ti;
1666
	sector_t max_sectors, max_size = 0;
M
Milan Broz 已提交
1667 1668

	if (unlikely(!map))
1669
		goto out;
M
Milan Broz 已提交
1670 1671

	ti = dm_table_find_target(map, bvm->bi_sector);
1672
	if (!dm_target_is_valid(ti))
M
Mikulas Patocka 已提交
1673
		goto out;
M
Milan Broz 已提交
1674 1675 1676 1677

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1678
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
1679
			  (sector_t) queue_max_sectors(q));
M
Milan Broz 已提交
1680
	max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size;
1681 1682 1683 1684 1685 1686 1687 1688 1689 1690

	/*
	 * FIXME: this stop-gap fix _must_ be cleaned up (by passing a sector_t
	 * to the targets' merge function since it holds sectors not bytes).
	 * Just doing this as an interim fix for stable@ because the more
	 * comprehensive cleanup of switching to sector_t will impact every
	 * DM target that implements a ->merge hook.
	 */
	if (max_size > INT_MAX)
		max_size = INT_MAX;
M
Milan Broz 已提交
1691 1692 1693 1694 1695 1696 1697

	/*
	 * 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)
1698
		max_size = ti->type->merge(ti, bvm, biovec, (int) max_size);
1699 1700
	/*
	 * If the target doesn't support merge method and some of the devices
1701 1702 1703 1704
	 * 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.
1705 1706 1707
	 */
	else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9)
		max_size = 0;
M
Milan Broz 已提交
1708

1709
out:
M
Mikulas Patocka 已提交
1710
	dm_put_live_table_fast(md);
M
Milan Broz 已提交
1711 1712 1713 1714 1715 1716 1717 1718 1719
	/*
	 * 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
1720 1721 1722 1723
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1724
static void dm_make_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1725
{
1726
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1727
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1728 1729
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
1730

M
Mikulas Patocka 已提交
1731
	map = dm_get_live_table(md, &srcu_idx);
L
Linus Torvalds 已提交
1732

1733
	generic_start_io_acct(rw, bio_sectors(bio), &dm_disk(md)->part0);
1734

1735 1736
	/* 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
1737
		dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
1738

1739 1740 1741
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1742
			bio_io_error(bio);
1743
		return;
L
Linus Torvalds 已提交
1744 1745
	}

M
Mikulas Patocka 已提交
1746 1747
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1748
	return;
1749 1750
}

M
Mikulas Patocka 已提交
1751
int dm_request_based(struct mapped_device *md)
1752 1753 1754 1755
{
	return blk_queue_stackable(md->queue);
}

1756
static void dm_dispatch_clone_request(struct request *clone, struct request *rq)
1757 1758 1759
{
	int r;

1760 1761
	if (blk_queue_io_stat(clone->q))
		clone->cmd_flags |= REQ_IO_STAT;
1762

1763 1764
	clone->start_time = jiffies;
	r = blk_insert_cloned_request(clone->q, clone);
1765
	if (r)
1766
		/* must complete clone in terms of original request */
1767 1768 1769
		dm_complete_request(rq, r);
}

1770 1771
static void setup_clone(struct request *clone, struct request *rq,
		        struct dm_rq_target_io *tio)
1772
{
1773
	blk_rq_prep_clone(clone, rq);
1774 1775
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;
1776
	tio->clone = clone;
1777 1778
}

K
Kiyoshi Ueda 已提交
1779
static struct request *clone_rq(struct request *rq, struct mapped_device *md,
1780
				struct dm_rq_target_io *tio, gfp_t gfp_mask)
1781
{
1782 1783 1784 1785 1786 1787
	/*
	 * Do not allocate a clone if tio->clone was already set
	 * (see: dm_mq_queue_rq).
	 */
	bool alloc_clone = !tio->clone;
	struct request *clone;
1788

1789 1790 1791 1792 1793 1794
	if (alloc_clone) {
		clone = alloc_clone_request(md, gfp_mask);
		if (!clone)
			return NULL;
	} else
		clone = tio->clone;
1795 1796

	blk_rq_init(NULL, clone);
1797
	setup_clone(clone, rq, tio);
1798 1799 1800 1801

	return clone;
}

1802 1803
static void map_tio_request(struct kthread_work *work);

1804 1805 1806 1807 1808 1809 1810 1811 1812
static void init_tio(struct dm_rq_target_io *tio, struct request *rq,
		     struct mapped_device *md)
{
	tio->md = md;
	tio->ti = NULL;
	tio->clone = NULL;
	tio->orig = rq;
	tio->error = 0;
	memset(&tio->info, 0, sizeof(tio->info));
1813 1814
	if (md->kworker_task)
		init_kthread_work(&tio->work, map_tio_request);
1815 1816
}

1817 1818
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
1819 1820
{
	struct dm_rq_target_io *tio;
1821 1822
	int srcu_idx;
	struct dm_table *table;
K
Kiyoshi Ueda 已提交
1823 1824 1825 1826 1827

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

1828
	init_tio(tio, rq, md);
K
Kiyoshi Ueda 已提交
1829

1830 1831 1832 1833 1834 1835 1836
	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
1837
	}
1838
	dm_put_live_table(md, srcu_idx);
K
Kiyoshi Ueda 已提交
1839

1840
	return tio;
K
Kiyoshi Ueda 已提交
1841 1842
}

1843 1844 1845 1846 1847 1848
/*
 * Called with the queue lock held.
 */
static int dm_prep_fn(struct request_queue *q, struct request *rq)
{
	struct mapped_device *md = q->queuedata;
1849
	struct dm_rq_target_io *tio;
1850 1851 1852 1853 1854 1855

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

1856 1857
	tio = prep_tio(rq, md, GFP_ATOMIC);
	if (!tio)
1858 1859
		return BLKPREP_DEFER;

1860
	rq->special = tio;
1861 1862 1863 1864 1865
	rq->cmd_flags |= REQ_DONTPREP;

	return BLKPREP_OK;
}

1866 1867
/*
 * Returns:
1868 1869 1870
 * 0                : the request has been processed
 * DM_MAPIO_REQUEUE : the original request needs to be requeued
 * < 0              : the request was completed due to failure
1871
 */
1872
static int map_request(struct dm_rq_target_io *tio, struct request *rq,
1873
		       struct mapped_device *md)
1874
{
1875
	int r;
1876
	struct dm_target *ti = tio->ti;
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;
		}
1889 1890
		if (r != DM_MAPIO_REMAPPED)
			return r;
1891
		setup_clone(clone, rq, tio);
1892
	}
1893 1894 1895 1896 1897 1898 1899

	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 */
1900
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
1901 1902
				     blk_rq_pos(rq));
		dm_dispatch_clone_request(clone, rq);
1903 1904 1905
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
1906
		dm_requeue_original_request(md, tio->orig);
1907 1908 1909 1910 1911 1912 1913 1914
		break;
	default:
		if (r > 0) {
			DMWARN("unimplemented target map return value: %d", r);
			BUG();
		}

		/* The target wants to complete the I/O */
1915
		dm_kill_unmapped_request(rq, r);
1916
		return r;
1917
	}
1918

1919
	return 0;
1920 1921
}

1922
static void map_tio_request(struct kthread_work *work)
1923
{
1924
	struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work);
1925 1926
	struct request *rq = tio->orig;
	struct mapped_device *md = tio->md;
1927

1928
	if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE)
1929
		dm_requeue_original_request(md, rq);
1930 1931
}

1932
static void dm_start_request(struct mapped_device *md, struct request *orig)
1933
{
1934 1935 1936 1937
	if (!orig->q->mq_ops)
		blk_start_request(orig);
	else
		blk_mq_start_request(orig);
1938
	atomic_inc(&md->pending[rq_data_dir(orig)]);
1939

1940 1941 1942 1943 1944
	if (md->seq_rq_merge_deadline_usecs) {
		md->last_rq_pos = rq_end_sector(orig);
		md->last_rq_rw = rq_data_dir(orig);
		md->last_rq_start_time = ktime_get();
	}
1945

1946 1947 1948 1949 1950 1951 1952 1953 1954 1955
	/*
	 * 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);
}

1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
#define MAX_SEQ_RQ_MERGE_DEADLINE_USECS 100000

ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf)
{
	return sprintf(buf, "%u\n", md->seq_rq_merge_deadline_usecs);
}

ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md,
						     const char *buf, size_t count)
{
	unsigned deadline;

1968
	if (!dm_request_based(md) || md->use_blk_mq)
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
		return count;

	if (kstrtouint(buf, 10, &deadline))
		return -EINVAL;

	if (deadline > MAX_SEQ_RQ_MERGE_DEADLINE_USECS)
		deadline = MAX_SEQ_RQ_MERGE_DEADLINE_USECS;

	md->seq_rq_merge_deadline_usecs = deadline;

	return count;
}

static bool dm_request_peeked_before_merge_deadline(struct mapped_device *md)
{
	ktime_t kt_deadline;

	if (!md->seq_rq_merge_deadline_usecs)
		return false;

	kt_deadline = ns_to_ktime((u64)md->seq_rq_merge_deadline_usecs * NSEC_PER_USEC);
	kt_deadline = ktime_add_safe(md->last_rq_start_time, kt_deadline);

	return !ktime_after(ktime_get(), kt_deadline);
}

1995 1996 1997 1998 1999 2000 2001
/*
 * 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
2002 2003
	int srcu_idx;
	struct dm_table *map = dm_get_live_table(md, &srcu_idx);
2004
	struct dm_target *ti;
2005
	struct request *rq;
2006
	struct dm_rq_target_io *tio;
2007
	sector_t pos;
2008 2009

	/*
2010 2011 2012 2013
	 * 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().
2014
	 */
J
Jens Axboe 已提交
2015
	while (!blk_queue_stopped(q)) {
2016 2017
		rq = blk_peek_request(q);
		if (!rq)
2018
			goto out;
2019

2020 2021 2022 2023 2024 2025
		/* 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);
2026 2027
		if (!dm_target_is_valid(ti)) {
			/*
2028
			 * Must perform setup, that rq_completed() requires,
2029 2030 2031
			 * before calling dm_kill_unmapped_request
			 */
			DMERR_LIMIT("request attempted access beyond the end of device");
2032 2033
			dm_start_request(md, rq);
			dm_kill_unmapped_request(rq, -EIO);
2034 2035
			continue;
		}
2036

2037 2038
		if (dm_request_peeked_before_merge_deadline(md) &&
		    md_in_flight(md) && rq->bio && rq->bio->bi_vcnt == 1 &&
2039 2040 2041
		    md->last_rq_pos == pos && md->last_rq_rw == rq_data_dir(rq))
			goto delay_and_out;

2042
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
2043
			goto delay_and_out;
2044

2045
		dm_start_request(md, rq);
2046

2047
		tio = tio_from_request(rq);
2048 2049 2050
		/* Establish tio->ti before queuing work (map_tio_request) */
		tio->ti = ti;
		queue_kthread_work(&md->kworker, &tio->work);
2051
		BUG_ON(!irqs_disabled());
2052 2053 2054 2055
	}

	goto out;

J
Jens Axboe 已提交
2056
delay_and_out:
2057
	blk_delay_queue(q, HZ / 100);
2058
out:
M
Mikulas Patocka 已提交
2059
	dm_put_live_table(md, srcu_idx);
2060 2061
}

L
Linus Torvalds 已提交
2062 2063
static int dm_any_congested(void *congested_data, int bdi_bits)
{
2064 2065 2066
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
2067

2068
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mikulas Patocka 已提交
2069
		map = dm_get_live_table_fast(md);
2070
		if (map) {
2071 2072 2073 2074 2075 2076 2077 2078 2079
			/*
			 * 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);
2080
		}
M
Mikulas Patocka 已提交
2081
		dm_put_live_table_fast(md);
2082 2083
	}

L
Linus Torvalds 已提交
2084 2085 2086 2087 2088 2089
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
2090
static void free_minor(int minor)
L
Linus Torvalds 已提交
2091
{
2092
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2093
	idr_remove(&_minor_idr, minor);
2094
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2095 2096 2097 2098 2099
}

/*
 * See if the device with a specific minor # is free.
 */
2100
static int specific_minor(int minor)
L
Linus Torvalds 已提交
2101
{
T
Tejun Heo 已提交
2102
	int r;
L
Linus Torvalds 已提交
2103 2104 2105 2106

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

T
Tejun Heo 已提交
2107
	idr_preload(GFP_KERNEL);
2108
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2109

T
Tejun Heo 已提交
2110
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
2111

2112
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
2113 2114 2115 2116
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
2117 2118
}

2119
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
2120
{
T
Tejun Heo 已提交
2121
	int r;
J
Jeff Mahoney 已提交
2122

T
Tejun Heo 已提交
2123
	idr_preload(GFP_KERNEL);
2124
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2125

T
Tejun Heo 已提交
2126
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
2127

2128
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
2129 2130 2131 2132 2133
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
2134 2135
}

2136
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
2137

2138 2139
static void dm_wq_work(struct work_struct *work);

2140 2141 2142 2143
static void dm_init_md_queue(struct mapped_device *md)
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
2144
	 * devices.  The type of this dm device may not have been decided yet.
2145 2146 2147 2148 2149 2150 2151
	 * 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);
2152
}
2153

2154 2155
static void dm_init_old_md_queue(struct mapped_device *md)
{
2156
	md->use_blk_mq = false;
2157 2158 2159 2160 2161
	dm_init_md_queue(md);

	/*
	 * Initialize aspects of queue that aren't relevant for blk-mq
	 */
2162 2163 2164
	md->queue->queuedata = md;
	md->queue->backing_dev_info.congested_fn = dm_any_congested;
	md->queue->backing_dev_info.congested_data = md;
2165

2166 2167 2168
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
}

2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202
static void cleanup_mapped_device(struct mapped_device *md)
{
	cleanup_srcu_struct(&md->io_barrier);

	if (md->wq)
		destroy_workqueue(md->wq);
	if (md->kworker_task)
		kthread_stop(md->kworker_task);
	if (md->io_pool)
		mempool_destroy(md->io_pool);
	if (md->rq_pool)
		mempool_destroy(md->rq_pool);
	if (md->bs)
		bioset_free(md->bs);

	if (md->disk) {
		spin_lock(&_minor_lock);
		md->disk->private_data = NULL;
		spin_unlock(&_minor_lock);
		if (blk_get_integrity(md->disk))
			blk_integrity_unregister(md->disk);
		del_gendisk(md->disk);
		put_disk(md->disk);
	}

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

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

L
Linus Torvalds 已提交
2203 2204 2205
/*
 * Allocate and initialise a blank device with a given minor.
 */
2206
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
2207 2208
{
	int r;
2209
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
2210
	void *old_md;
L
Linus Torvalds 已提交
2211 2212 2213 2214 2215 2216

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

2217
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
2218
		goto bad_module_get;
2219

L
Linus Torvalds 已提交
2220
	/* get a minor number for the dev */
2221
	if (minor == DM_ANY_MINOR)
2222
		r = next_free_minor(&minor);
2223
	else
2224
		r = specific_minor(minor);
L
Linus Torvalds 已提交
2225
	if (r < 0)
M
Milan Broz 已提交
2226
		goto bad_minor;
L
Linus Torvalds 已提交
2227

M
Mikulas Patocka 已提交
2228 2229 2230 2231
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

2232
	md->use_blk_mq = use_blk_mq;
2233
	md->type = DM_TYPE_NONE;
2234
	mutex_init(&md->suspend_lock);
2235
	mutex_init(&md->type_lock);
2236
	mutex_init(&md->table_devices_lock);
2237
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
2238
	atomic_set(&md->holders, 1);
2239
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
2240
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
2241 2242
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
2243
	INIT_LIST_HEAD(&md->table_devices);
M
Mike Anderson 已提交
2244
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
2245

2246
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
2247
	if (!md->queue)
2248
		goto bad;
L
Linus Torvalds 已提交
2249

2250
	dm_init_md_queue(md);
S
Stefan Bader 已提交
2251

L
Linus Torvalds 已提交
2252 2253
	md->disk = alloc_disk(1);
	if (!md->disk)
2254
		goto bad;
L
Linus Torvalds 已提交
2255

2256 2257
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
2258
	init_waitqueue_head(&md->wait);
2259
	INIT_WORK(&md->work, dm_wq_work);
2260
	init_waitqueue_head(&md->eventq);
2261
	init_completion(&md->kobj_holder.completion);
2262
	md->kworker_task = NULL;
2263

L
Linus Torvalds 已提交
2264 2265 2266 2267 2268 2269 2270
	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
2271
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
2272

T
Tejun Heo 已提交
2273
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
2274
	if (!md->wq)
2275
		goto bad;
2276

M
Mikulas Patocka 已提交
2277 2278
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
2279
		goto bad;
M
Mikulas Patocka 已提交
2280

2281 2282 2283 2284
	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
2285 2286
	dm_stats_init(&md->stats);

2287
	/* Populate the mapping, nobody knows we exist yet */
2288
	spin_lock(&_minor_lock);
2289
	old_md = idr_replace(&_minor_idr, md, minor);
2290
	spin_unlock(&_minor_lock);
2291 2292 2293

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
2294 2295
	return md;

2296 2297
bad:
	cleanup_mapped_device(md);
M
Mikulas Patocka 已提交
2298
bad_io_barrier:
L
Linus Torvalds 已提交
2299
	free_minor(minor);
M
Milan Broz 已提交
2300
bad_minor:
2301
	module_put(THIS_MODULE);
M
Milan Broz 已提交
2302
bad_module_get:
L
Linus Torvalds 已提交
2303 2304 2305 2306
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
2307 2308
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
2309 2310
static void free_dev(struct mapped_device *md)
{
2311
	int minor = MINOR(disk_devt(md->disk));
2312

M
Mikulas Patocka 已提交
2313
	unlock_fs(md);
2314

2315 2316 2317
	cleanup_mapped_device(md);
	if (md->use_blk_mq)
		blk_mq_free_tag_set(&md->tag_set);
2318

2319
	free_table_devices(&md->table_devices);
2320 2321 2322
	dm_stats_cleanup(&md->stats);
	free_minor(minor);

2323
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
2324 2325 2326
	kfree(md);
}

2327 2328 2329 2330 2331 2332 2333 2334
static unsigned filter_md_type(unsigned type, struct mapped_device *md)
{
	if (type == DM_TYPE_BIO_BASED)
		return type;

	return !md->use_blk_mq ? DM_TYPE_REQUEST_BASED : DM_TYPE_MQ_REQUEST_BASED;
}

K
Kiyoshi Ueda 已提交
2335 2336
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
2337
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
2338

2339 2340 2341
	switch (filter_md_type(dm_table_get_type(t), md)) {
	case DM_TYPE_BIO_BASED:
		if (md->bs && md->io_pool) {
2342
			/*
2343
			 * This bio-based md already has necessary mempools.
2344 2345 2346 2347 2348 2349
			 * 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;
2350
			goto out;
2351
		}
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
		break;
	case DM_TYPE_REQUEST_BASED:
		if (md->rq_pool && md->io_pool)
			/*
			 * This request-based md already has necessary mempools.
			 */
			goto out;
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
		BUG_ON(p); /* No mempools needed */
		return;
M
Mikulas Patocka 已提交
2363
	}
K
Kiyoshi Ueda 已提交
2364

2365 2366
	BUG_ON(!p || md->io_pool || md->rq_pool || md->bs);

K
Kiyoshi Ueda 已提交
2367 2368
	md->io_pool = p->io_pool;
	p->io_pool = NULL;
2369 2370
	md->rq_pool = p->rq_pool;
	p->rq_pool = NULL;
K
Kiyoshi Ueda 已提交
2371 2372 2373
	md->bs = p->bs;
	p->bs = NULL;
out:
2374
	/* mempool bind completed, no longer need any mempools in the table */
K
Kiyoshi Ueda 已提交
2375 2376 2377
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
2378 2379 2380 2381 2382
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2383 2384
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2385 2386
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2387 2388 2389 2390
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2391
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2392

L
Linus Torvalds 已提交
2393 2394 2395 2396
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2397 2398 2399
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2400
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2401
{
2402
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2403

2404
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2405 2406
}

2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420
/*
 * 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;

2421
	if (q->make_request_fn == dm_make_request) {
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459
		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;
}

2460 2461 2462 2463 2464
/*
 * 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
2465
{
2466
	struct dm_table *old_map;
2467
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2468
	sector_t size;
2469
	int merge_is_optional;
L
Linus Torvalds 已提交
2470 2471

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2472 2473 2474 2475

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
2476
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
2477 2478
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
2479
	__set_size(md, size);
2480

2481 2482
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2483 2484 2485 2486 2487 2488 2489
	/*
	 * 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.
	 */
2490
	if (dm_table_request_based(t))
K
Kiyoshi Ueda 已提交
2491 2492 2493 2494
		stop_queue(q);

	__bind_mempools(md, t);

2495 2496
	merge_is_optional = dm_table_merge_is_optional(t);

2497
	old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
M
Mikulas Patocka 已提交
2498
	rcu_assign_pointer(md->map, t);
2499 2500
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2501
	dm_table_set_restrictions(t, q, limits);
2502 2503 2504 2505
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
2506 2507
	if (old_map)
		dm_sync_table(md);
L
Linus Torvalds 已提交
2508

2509
	return old_map;
L
Linus Torvalds 已提交
2510 2511
}

2512 2513 2514 2515
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2516
{
2517
	struct dm_table *map = rcu_dereference_protected(md->map, 1);
L
Linus Torvalds 已提交
2518 2519

	if (!map)
2520
		return NULL;
L
Linus Torvalds 已提交
2521 2522

	dm_table_event_callback(map, NULL, NULL);
2523
	RCU_INIT_POINTER(md->map, NULL);
M
Mikulas Patocka 已提交
2524
	dm_sync_table(md);
2525 2526

	return map;
L
Linus Torvalds 已提交
2527 2528 2529 2530 2531
}

/*
 * Constructor for a new device.
 */
2532
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2533 2534 2535
{
	struct mapped_device *md;

2536
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2537 2538 2539
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2540 2541
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2542 2543 2544 2545
	*result = md;
	return 0;
}

2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561
/*
 * 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)
{
2562
	BUG_ON(!mutex_is_locked(&md->type_lock));
2563 2564 2565 2566 2567
	md->type = type;
}

unsigned dm_get_md_type(struct mapped_device *md)
{
2568
	BUG_ON(!mutex_is_locked(&md->type_lock));
2569 2570 2571
	return md->type;
}

2572 2573 2574 2575 2576
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
/*
 * 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);

2588 2589 2590 2591 2592 2593 2594 2595
static void init_rq_based_worker_thread(struct mapped_device *md)
{
	/* 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));
}

2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
/*
 * 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;

	/* Fully initialize the queue */
	q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL);
	if (!q)
2606
		return -EINVAL;
2607

2608 2609 2610
	/* disable dm_request_fn's merge heuristic by default */
	md->seq_rq_merge_deadline_usecs = 0;

2611
	md->queue = q;
2612
	dm_init_old_md_queue(md);
2613 2614 2615
	blk_queue_softirq_done(md->queue, dm_softirq_done);
	blk_queue_prep_rq(md->queue, dm_prep_fn);

2616
	init_rq_based_worker_thread(md);
2617

2618 2619
	elv_register_queue(md->queue);

2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675
	return 0;
}

static int dm_mq_init_request(void *data, struct request *rq,
			      unsigned int hctx_idx, unsigned int request_idx,
			      unsigned int numa_node)
{
	struct mapped_device *md = data;
	struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);

	/*
	 * Must initialize md member of tio, otherwise it won't
	 * be available in dm_mq_queue_rq.
	 */
	tio->md = md;

	return 0;
}

static int dm_mq_queue_rq(struct blk_mq_hw_ctx *hctx,
			  const struct blk_mq_queue_data *bd)
{
	struct request *rq = bd->rq;
	struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);
	struct mapped_device *md = tio->md;
	int srcu_idx;
	struct dm_table *map = dm_get_live_table(md, &srcu_idx);
	struct dm_target *ti;
	sector_t pos;

	/* 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);
	if (!dm_target_is_valid(ti)) {
		dm_put_live_table(md, srcu_idx);
		DMERR_LIMIT("request attempted access beyond the end of device");
		/*
		 * Must perform setup, that rq_completed() requires,
		 * before returning BLK_MQ_RQ_QUEUE_ERROR
		 */
		dm_start_request(md, rq);
		return BLK_MQ_RQ_QUEUE_ERROR;
	}
	dm_put_live_table(md, srcu_idx);

	if (ti->type->busy && ti->type->busy(ti))
		return BLK_MQ_RQ_QUEUE_BUSY;

	dm_start_request(md, rq);

	/* Init tio using md established in .init_request */
	init_tio(tio, rq, md);

2676 2677 2678 2679
	/*
	 * Establish tio->ti before queuing work (map_tio_request)
	 * or making direct call to map_request().
	 */
2680
	tio->ti = ti;
2681 2682 2683 2684 2685

	/* Clone the request if underlying devices aren't blk-mq */
	if (dm_table_get_type(map) == DM_TYPE_REQUEST_BASED) {
		/* clone request is allocated at the end of the pdu */
		tio->clone = (void *)blk_mq_rq_to_pdu(rq) + sizeof(struct dm_rq_target_io);
2686
		(void) clone_rq(rq, md, tio, GFP_ATOMIC);
2687 2688 2689
		queue_kthread_work(&md->kworker, &tio->work);
	} else {
		/* Direct call is fine since .queue_rq allows allocations */
2690 2691 2692 2693 2694
		if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE) {
			/* Undo dm_start_request() before requeuing */
			rq_completed(md, rq_data_dir(rq), false);
			return BLK_MQ_RQ_QUEUE_BUSY;
		}
2695
	}
2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708

	return BLK_MQ_RQ_QUEUE_OK;
}

static struct blk_mq_ops dm_mq_ops = {
	.queue_rq = dm_mq_queue_rq,
	.map_queue = blk_mq_map_queue,
	.complete = dm_softirq_done,
	.init_request = dm_mq_init_request,
};

static int dm_init_request_based_blk_mq_queue(struct mapped_device *md)
{
2709
	unsigned md_type = dm_get_md_type(md);
2710 2711 2712 2713 2714 2715 2716 2717 2718
	struct request_queue *q;
	int err;

	memset(&md->tag_set, 0, sizeof(md->tag_set));
	md->tag_set.ops = &dm_mq_ops;
	md->tag_set.queue_depth = BLKDEV_MAX_RQ;
	md->tag_set.numa_node = NUMA_NO_NODE;
	md->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
	md->tag_set.nr_hw_queues = 1;
2719 2720 2721 2722 2723
	if (md_type == DM_TYPE_REQUEST_BASED) {
		/* make the memory for non-blk-mq clone part of the pdu */
		md->tag_set.cmd_size = sizeof(struct dm_rq_target_io) + sizeof(struct request);
	} else
		md->tag_set.cmd_size = sizeof(struct dm_rq_target_io);
2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740
	md->tag_set.driver_data = md;

	err = blk_mq_alloc_tag_set(&md->tag_set);
	if (err)
		return err;

	q = blk_mq_init_allocated_queue(&md->tag_set, md->queue);
	if (IS_ERR(q)) {
		err = PTR_ERR(q);
		goto out_tag_set;
	}
	md->queue = q;
	dm_init_md_queue(md);

	/* backfill 'mq' sysfs registration normally done in blk_register_queue */
	blk_mq_register_disk(md->disk);

2741 2742
	if (md_type == DM_TYPE_REQUEST_BASED)
		init_rq_based_worker_thread(md);
2743 2744 2745 2746 2747 2748

	return 0;

out_tag_set:
	blk_mq_free_tag_set(&md->tag_set);
	return err;
2749 2750 2751 2752 2753 2754 2755
}

/*
 * Setup the DM device's queue based on md's type
 */
int dm_setup_md_queue(struct mapped_device *md)
{
2756
	int r;
2757
	unsigned md_type = filter_md_type(dm_get_md_type(md), md);
2758 2759 2760 2761 2762

	switch (md_type) {
	case DM_TYPE_REQUEST_BASED:
		r = dm_init_request_based_queue(md);
		if (r) {
2763
			DMWARN("Cannot initialize queue for request-based mapped device");
2764
			return r;
2765
		}
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775
		break;
	case DM_TYPE_MQ_REQUEST_BASED:
		r = dm_init_request_based_blk_mq_queue(md);
		if (r) {
			DMWARN("Cannot initialize queue for request-based blk-mq mapped device");
			return r;
		}
		break;
	case DM_TYPE_BIO_BASED:
		dm_init_old_md_queue(md);
2776 2777
		blk_queue_make_request(md->queue, dm_make_request);
		blk_queue_merge_bvec(md->queue, dm_merge_bvec);
2778
		break;
2779 2780 2781 2782 2783
	}

	return 0;
}

2784
struct mapped_device *dm_get_md(dev_t dev)
L
Linus Torvalds 已提交
2785 2786 2787 2788 2789 2790 2791
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2792
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2793 2794

	md = idr_find(&_minor_idr, minor);
2795 2796 2797 2798 2799 2800 2801 2802 2803
	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
2804
	}
L
Linus Torvalds 已提交
2805

J
Jeff Mahoney 已提交
2806
out:
2807
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2808

2809 2810
	return md;
}
A
Alasdair G Kergon 已提交
2811
EXPORT_SYMBOL_GPL(dm_get_md);
2812

A
Alasdair G Kergon 已提交
2813
void *dm_get_mdptr(struct mapped_device *md)
2814
{
A
Alasdair G Kergon 已提交
2815
	return md->interface_ptr;
L
Linus Torvalds 已提交
2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
}

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);
2826
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2827 2828
}

2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841
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);

2842 2843 2844 2845 2846 2847
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2848
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2849
{
M
Mike Anderson 已提交
2850
	struct dm_table *map;
M
Mikulas Patocka 已提交
2851
	int srcu_idx;
L
Linus Torvalds 已提交
2852

2853
	might_sleep();
J
Jeff Mahoney 已提交
2854

M
Mikulas Patocka 已提交
2855
	map = dm_get_live_table(md, &srcu_idx);
2856 2857

	spin_lock(&_minor_lock);
2858 2859 2860 2861
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);

2862
	if (dm_request_based(md) && md->kworker_task)
2863 2864
		flush_kthread_worker(&md->kworker);

2865 2866 2867 2868 2869
	/*
	 * Take suspend_lock so that presuspend and postsuspend methods
	 * do not race with internal suspend.
	 */
	mutex_lock(&md->suspend_lock);
2870 2871 2872
	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2873
	}
2874
	mutex_unlock(&md->suspend_lock);
2875

M
Mikulas Patocka 已提交
2876 2877 2878
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);

2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909
	/*
	 * 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
2910
}
E
Edward Goggin 已提交
2911
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2912

2913
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2914 2915
{
	int r = 0;
2916 2917 2918
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2919 2920

	while (1) {
2921
		set_current_state(interruptible);
2922

2923
		if (!md_in_flight(md))
2924 2925
			break;

2926 2927
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2928 2929 2930 2931 2932 2933 2934 2935
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2936 2937
	remove_wait_queue(&md->wait, &wait);

2938 2939 2940
	return r;
}

L
Linus Torvalds 已提交
2941 2942 2943
/*
 * Process the deferred bios
 */
2944
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2945
{
2946 2947
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2948
	struct bio *c;
M
Mikulas Patocka 已提交
2949 2950
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2951

M
Mikulas Patocka 已提交
2952
	map = dm_get_live_table(md, &srcu_idx);
2953

2954
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2955 2956 2957 2958
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2959
		if (!c)
A
Alasdair G Kergon 已提交
2960
			break;
2961

K
Kiyoshi Ueda 已提交
2962 2963
		if (dm_request_based(md))
			generic_make_request(c);
2964
		else
M
Mikulas Patocka 已提交
2965
			__split_and_process_bio(md, map, c);
2966
	}
M
Milan Broz 已提交
2967

M
Mikulas Patocka 已提交
2968
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2969 2970
}

2971
static void dm_queue_flush(struct mapped_device *md)
2972
{
2973
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2974
	smp_mb__after_atomic();
2975
	queue_work(md->wq, &md->work);
2976 2977
}

L
Linus Torvalds 已提交
2978
/*
2979
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2980
 */
2981
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2982
{
2983
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2984
	struct queue_limits limits;
2985
	int r;
L
Linus Torvalds 已提交
2986

2987
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2988 2989

	/* device must be suspended */
2990
	if (!dm_suspended_md(md))
2991
		goto out;
L
Linus Torvalds 已提交
2992

2993 2994 2995 2996 2997 2998 2999
	/*
	 * 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
3000
		live_map = dm_get_live_table_fast(md);
3001 3002
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
3003
		dm_put_live_table_fast(md);
3004 3005
	}

3006 3007 3008 3009 3010 3011
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
3012
	}
3013

3014
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
3015

3016
out:
3017
	mutex_unlock(&md->suspend_lock);
3018
	return map;
L
Linus Torvalds 已提交
3019 3020 3021 3022 3023 3024
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
3025
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
3026
{
3027
	int r;
L
Linus Torvalds 已提交
3028 3029

	WARN_ON(md->frozen_sb);
3030

3031
	md->frozen_sb = freeze_bdev(md->bdev);
3032
	if (IS_ERR(md->frozen_sb)) {
3033
		r = PTR_ERR(md->frozen_sb);
3034 3035
		md->frozen_sb = NULL;
		return r;
3036 3037
	}

3038 3039
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
3040 3041 3042
	return 0;
}

3043
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
3044
{
3045 3046 3047
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

3048
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
3049
	md->frozen_sb = NULL;
3050
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
3051 3052 3053
}

/*
3054 3055 3056
 * 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.
3057
 *
3058
 * Caller must hold md->suspend_lock
3059
 */
3060 3061
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
			unsigned suspend_flags, int interruptible)
L
Linus Torvalds 已提交
3062
{
3063 3064 3065
	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
3066

3067 3068 3069 3070 3071 3072 3073
	/*
	 * 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);

3074 3075 3076 3077
	/*
	 * This gets reverted if there's an error later and the targets
	 * provide the .presuspend_undo hook.
	 */
3078 3079
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
3080
	/*
K
Kiyoshi Ueda 已提交
3081 3082 3083 3084
	 * 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
3085 3086 3087
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
3088 3089
		if (r) {
			dm_table_presuspend_undo_targets(map);
3090
			return r;
3091
		}
3092
	}
L
Linus Torvalds 已提交
3093 3094

	/*
3095 3096 3097 3098 3099 3100 3101
	 * 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
3102 3103 3104
	 * __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
3105
	 */
3106
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
3107 3108
	if (map)
		synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
3109

3110
	/*
3111 3112
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
3113
	 */
3114
	if (dm_request_based(md)) {
K
Kiyoshi Ueda 已提交
3115
		stop_queue(md->queue);
3116 3117
		if (md->kworker_task)
			flush_kthread_worker(&md->kworker);
3118
	}
3119

3120 3121
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
3122
	/*
3123 3124 3125
	 * 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
3126
	 */
3127
	r = dm_wait_for_completion(md, interruptible);
L
Linus Torvalds 已提交
3128

3129
	if (noflush)
3130
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
3131 3132
	if (map)
		synchronize_srcu(&md->io_barrier);
3133

L
Linus Torvalds 已提交
3134
	/* were we interrupted ? */
3135
	if (r < 0) {
3136
		dm_queue_flush(md);
M
Milan Broz 已提交
3137

3138
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
3139
			start_queue(md->queue);
3140

3141
		unlock_fs(md);
3142
		dm_table_presuspend_undo_targets(map);
3143
		/* pushback list is already flushed, so skip flush */
3144
	}
L
Linus Torvalds 已提交
3145

3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
	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;
	}

3187
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
3188 3189 3190 3191

	r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE);
	if (r)
		goto out_unlock;
3192

3193
	set_bit(DMF_SUSPENDED, &md->flags);
3194

3195 3196
	dm_table_postsuspend_targets(map);

3197
out_unlock:
3198
	mutex_unlock(&md->suspend_lock);
3199
	return r;
L
Linus Torvalds 已提交
3200 3201
}

3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224
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
3225 3226
int dm_resume(struct mapped_device *md)
{
3227 3228
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
3229

3230 3231 3232
retry:
	mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);

3233
	if (!dm_suspended_md(md))
3234 3235
		goto out;

3236 3237 3238 3239 3240 3241 3242 3243 3244
	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;
	}

3245
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
3246
	if (!map || !dm_table_get_size(map))
3247
		goto out;
L
Linus Torvalds 已提交
3248

3249
	r = __dm_resume(md, map);
3250 3251
	if (r)
		goto out;
3252 3253 3254

	clear_bit(DMF_SUSPENDED, &md->flags);

3255 3256
	r = 0;
out:
3257
	mutex_unlock(&md->suspend_lock);
3258

3259
	return r;
L
Linus Torvalds 已提交
3260 3261
}

M
Mikulas Patocka 已提交
3262 3263 3264 3265 3266 3267
/*
 * 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.
 */

3268
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
M
Mikulas Patocka 已提交
3269
{
3270 3271
	struct dm_table *map = NULL;

3272
	if (md->internal_suspend_count++)
3273 3274 3275 3276 3277 3278 3279
		return; /* nested internal suspend */

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

3280
	map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296

	/*
	 * 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)
{
3297 3298 3299
	BUG_ON(!md->internal_suspend_count);

	if (--md->internal_suspend_count)
3300 3301
		return; /* resume from nested internal suspend */

M
Mikulas Patocka 已提交
3302
	if (dm_suspended_md(md))
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
		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
3342 3343 3344 3345 3346 3347 3348
		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);
}
3349
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
M
Mikulas Patocka 已提交
3350

3351
void dm_internal_resume_fast(struct mapped_device *md)
M
Mikulas Patocka 已提交
3352
{
3353
	if (dm_suspended_md(md) || dm_suspended_internally_md(md))
M
Mikulas Patocka 已提交
3354 3355 3356 3357 3358 3359 3360
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}
3361
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
M
Mikulas Patocka 已提交
3362

L
Linus Torvalds 已提交
3363 3364 3365
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
3366
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
3367
		       unsigned cookie)
3368
{
M
Milan Broz 已提交
3369 3370 3371 3372
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
3373
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
3374 3375 3376
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
3377 3378
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
3379
	}
3380 3381
}

M
Mike Anderson 已提交
3382 3383 3384 3385 3386
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397
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
3398 3399 3400 3401 3402 3403 3404 3405 3406
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
3407 3408 3409 3410 3411 3412 3413 3414
/*
 * 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;
}
3415
EXPORT_SYMBOL_GPL(dm_disk);
L
Linus Torvalds 已提交
3416

M
Milan Broz 已提交
3417 3418
struct kobject *dm_kobject(struct mapped_device *md)
{
3419
	return &md->kobj_holder.kobj;
M
Milan Broz 已提交
3420 3421 3422 3423 3424 3425
}

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

3426
	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
M
Milan Broz 已提交
3427

3428
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
3429
	    dm_deleting_md(md))
3430 3431
		return NULL;

M
Milan Broz 已提交
3432 3433 3434 3435
	dm_get(md);
	return md;
}

3436
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
3437 3438 3439 3440
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

3441 3442 3443 3444 3445
int dm_suspended_internally_md(struct mapped_device *md)
{
	return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}

M
Mikulas Patocka 已提交
3446 3447 3448 3449 3450
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}

3451 3452
int dm_suspended(struct dm_target *ti)
{
3453
	return dm_suspended_md(dm_table_get_md(ti->table));
3454 3455 3456
}
EXPORT_SYMBOL_GPL(dm_suspended);

3457 3458
int dm_noflush_suspending(struct dm_target *ti)
{
3459
	return __noflush_suspending(dm_table_get_md(ti->table));
3460 3461 3462
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

3463 3464
struct dm_md_mempools *dm_alloc_bio_mempools(unsigned integrity,
					     unsigned per_bio_data_size)
K
Kiyoshi Ueda 已提交
3465
{
3466 3467
	struct dm_md_mempools *pools;
	unsigned int pool_size = dm_get_reserved_bio_based_ios();
J
Jun'ichi Nomura 已提交
3468
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
3469

3470
	pools = kzalloc(sizeof(*pools), GFP_KERNEL);
K
Kiyoshi Ueda 已提交
3471
	if (!pools)
3472
		return ERR_PTR(-ENOMEM);
K
Kiyoshi Ueda 已提交
3473

3474 3475
	front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) +
		offsetof(struct dm_target_io, clone);
3476

3477 3478 3479
	pools->io_pool = mempool_create_slab_pool(pool_size, _io_cache);
	if (!pools->io_pool)
		goto out;
K
Kiyoshi Ueda 已提交
3480

J
Junichi Nomura 已提交
3481
	pools->bs = bioset_create_nobvec(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
3482
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
3483
		goto out;
K
Kiyoshi Ueda 已提交
3484

3485
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
3486
		goto out;
3487

K
Kiyoshi Ueda 已提交
3488
	return pools;
J
Jun'ichi Nomura 已提交
3489 3490
out:
	dm_free_md_mempools(pools);
3491
	return ERR_PTR(-ENOMEM);
3492 3493 3494 3495 3496
}

struct dm_md_mempools *dm_alloc_rq_mempools(struct mapped_device *md,
					    unsigned type)
{
3497
	unsigned int pool_size;
3498
	struct dm_md_mempools *pools;
K
Kiyoshi Ueda 已提交
3499

3500 3501 3502 3503
	if (filter_md_type(type, md) == DM_TYPE_MQ_REQUEST_BASED)
		return NULL; /* No mempools needed */

	pool_size = dm_get_reserved_rq_based_ios();
3504 3505
	pools = kzalloc(sizeof(*pools), GFP_KERNEL);
	if (!pools)
3506
		return ERR_PTR(-ENOMEM);
3507

3508 3509 3510
	pools->rq_pool = mempool_create_slab_pool(pool_size, _rq_cache);
	if (!pools->rq_pool)
		goto out;
K
Kiyoshi Ueda 已提交
3511

3512 3513 3514 3515 3516 3517 3518
	pools->io_pool = mempool_create_slab_pool(pool_size, _rq_tio_cache);
	if (!pools->io_pool)
		goto out;

	return pools;
out:
	dm_free_md_mempools(pools);
3519
	return ERR_PTR(-ENOMEM);
K
Kiyoshi Ueda 已提交
3520 3521 3522 3523 3524 3525 3526 3527 3528 3529
}

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

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

3530 3531 3532
	if (pools->rq_pool)
		mempool_destroy(pools->rq_pool);

K
Kiyoshi Ueda 已提交
3533 3534 3535 3536 3537 3538
	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

3539
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
3540 3541
	.open = dm_blk_open,
	.release = dm_blk_close,
3542
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
3543
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554
	.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");
3555

3556 3557 3558
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

3559 3560 3561
module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");

3562 3563 3564
module_param(use_blk_mq, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(use_blk_mq, "Use block multiqueue for request-based DM devices");

L
Linus Torvalds 已提交
3565 3566 3567
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");