Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (C) 2001, 2002 Sistina Software (UK) Limited. | |
784aae73 | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
1da177e4 LT |
4 | * |
5 | * This file is released under the GPL. | |
6 | */ | |
7 | ||
8 | #include "dm.h" | |
51e5b2bd | 9 | #include "dm-uevent.h" |
1da177e4 LT |
10 | |
11 | #include <linux/init.h> | |
12 | #include <linux/module.h> | |
48c9c27b | 13 | #include <linux/mutex.h> |
1da177e4 LT |
14 | #include <linux/moduleparam.h> |
15 | #include <linux/blkpg.h> | |
16 | #include <linux/bio.h> | |
17 | #include <linux/buffer_head.h> | |
18 | #include <linux/mempool.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/idr.h> | |
3ac51e74 | 21 | #include <linux/hdreg.h> |
55782138 LZ |
22 | |
23 | #include <trace/events/block.h> | |
1da177e4 | 24 | |
72d94861 AK |
25 | #define DM_MSG_PREFIX "core" |
26 | ||
1da177e4 LT |
27 | static const char *_name = DM_NAME; |
28 | ||
29 | static unsigned int major = 0; | |
30 | static unsigned int _major = 0; | |
31 | ||
f32c10b0 | 32 | static DEFINE_SPINLOCK(_minor_lock); |
1da177e4 | 33 | /* |
8fbf26ad | 34 | * For bio-based dm. |
1da177e4 LT |
35 | * One of these is allocated per bio. |
36 | */ | |
37 | struct dm_io { | |
38 | struct mapped_device *md; | |
39 | int error; | |
1da177e4 | 40 | atomic_t io_count; |
6ae2fa67 | 41 | struct bio *bio; |
3eaf840e | 42 | unsigned long start_time; |
1da177e4 LT |
43 | }; |
44 | ||
45 | /* | |
8fbf26ad | 46 | * For bio-based dm. |
1da177e4 LT |
47 | * One of these is allocated per target within a bio. Hopefully |
48 | * this will be simplified out one day. | |
49 | */ | |
028867ac | 50 | struct dm_target_io { |
1da177e4 LT |
51 | struct dm_io *io; |
52 | struct dm_target *ti; | |
53 | union map_info info; | |
54 | }; | |
55 | ||
8fbf26ad KU |
56 | /* |
57 | * For request-based dm. | |
58 | * One of these is allocated per request. | |
59 | */ | |
60 | struct dm_rq_target_io { | |
61 | struct mapped_device *md; | |
62 | struct dm_target *ti; | |
63 | struct request *orig, clone; | |
64 | int error; | |
65 | union map_info info; | |
66 | }; | |
67 | ||
68 | /* | |
69 | * For request-based dm. | |
70 | * One of these is allocated per bio. | |
71 | */ | |
72 | struct dm_rq_clone_bio_info { | |
73 | struct bio *orig; | |
74 | struct request *rq; | |
75 | }; | |
76 | ||
1da177e4 LT |
77 | union map_info *dm_get_mapinfo(struct bio *bio) |
78 | { | |
17b2f66f | 79 | if (bio && bio->bi_private) |
028867ac | 80 | return &((struct dm_target_io *)bio->bi_private)->info; |
17b2f66f | 81 | return NULL; |
1da177e4 LT |
82 | } |
83 | ||
ba61fdd1 JM |
84 | #define MINOR_ALLOCED ((void *)-1) |
85 | ||
1da177e4 LT |
86 | /* |
87 | * Bits for the md->flags field. | |
88 | */ | |
1eb787ec | 89 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
1da177e4 | 90 | #define DMF_SUSPENDED 1 |
aa8d7c2f | 91 | #define DMF_FROZEN 2 |
fba9f90e | 92 | #define DMF_FREEING 3 |
5c6bd75d | 93 | #define DMF_DELETING 4 |
2e93ccc1 | 94 | #define DMF_NOFLUSH_SUSPENDING 5 |
1eb787ec | 95 | #define DMF_QUEUE_IO_TO_THREAD 6 |
1da177e4 | 96 | |
304f3f6a MB |
97 | /* |
98 | * Work processed by per-device workqueue. | |
99 | */ | |
1da177e4 | 100 | struct mapped_device { |
2ca3310e | 101 | struct rw_semaphore io_lock; |
e61290a4 | 102 | struct mutex suspend_lock; |
1da177e4 LT |
103 | rwlock_t map_lock; |
104 | atomic_t holders; | |
5c6bd75d | 105 | atomic_t open_count; |
1da177e4 LT |
106 | |
107 | unsigned long flags; | |
108 | ||
165125e1 | 109 | struct request_queue *queue; |
1da177e4 | 110 | struct gendisk *disk; |
7e51f257 | 111 | char name[16]; |
1da177e4 LT |
112 | |
113 | void *interface_ptr; | |
114 | ||
115 | /* | |
116 | * A list of ios that arrived while we were suspended. | |
117 | */ | |
118 | atomic_t pending; | |
119 | wait_queue_head_t wait; | |
53d5914f | 120 | struct work_struct work; |
74859364 | 121 | struct bio_list deferred; |
022c2611 | 122 | spinlock_t deferred_lock; |
1da177e4 | 123 | |
af7e466a MP |
124 | /* |
125 | * An error from the barrier request currently being processed. | |
126 | */ | |
127 | int barrier_error; | |
128 | ||
304f3f6a MB |
129 | /* |
130 | * Processing queue (flush/barriers) | |
131 | */ | |
132 | struct workqueue_struct *wq; | |
133 | ||
1da177e4 LT |
134 | /* |
135 | * The current mapping. | |
136 | */ | |
137 | struct dm_table *map; | |
138 | ||
139 | /* | |
140 | * io objects are allocated from here. | |
141 | */ | |
142 | mempool_t *io_pool; | |
143 | mempool_t *tio_pool; | |
144 | ||
9faf400f SB |
145 | struct bio_set *bs; |
146 | ||
1da177e4 LT |
147 | /* |
148 | * Event handling. | |
149 | */ | |
150 | atomic_t event_nr; | |
151 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
152 | atomic_t uevent_seq; |
153 | struct list_head uevent_list; | |
154 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
155 | |
156 | /* | |
157 | * freeze/thaw support require holding onto a super block | |
158 | */ | |
159 | struct super_block *frozen_sb; | |
e39e2e95 | 160 | struct block_device *suspended_bdev; |
3ac51e74 DW |
161 | |
162 | /* forced geometry settings */ | |
163 | struct hd_geometry geometry; | |
784aae73 MB |
164 | |
165 | /* sysfs handle */ | |
166 | struct kobject kobj; | |
1da177e4 LT |
167 | }; |
168 | ||
169 | #define MIN_IOS 256 | |
e18b890b CL |
170 | static struct kmem_cache *_io_cache; |
171 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
172 | static struct kmem_cache *_rq_tio_cache; |
173 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 174 | |
1da177e4 LT |
175 | static int __init local_init(void) |
176 | { | |
51157b4a | 177 | int r = -ENOMEM; |
1da177e4 | 178 | |
1da177e4 | 179 | /* allocate a slab for the dm_ios */ |
028867ac | 180 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 181 | if (!_io_cache) |
51157b4a | 182 | return r; |
1da177e4 LT |
183 | |
184 | /* allocate a slab for the target ios */ | |
028867ac | 185 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
186 | if (!_tio_cache) |
187 | goto out_free_io_cache; | |
1da177e4 | 188 | |
8fbf26ad KU |
189 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
190 | if (!_rq_tio_cache) | |
191 | goto out_free_tio_cache; | |
192 | ||
193 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
194 | if (!_rq_bio_info_cache) | |
195 | goto out_free_rq_tio_cache; | |
196 | ||
51e5b2bd | 197 | r = dm_uevent_init(); |
51157b4a | 198 | if (r) |
8fbf26ad | 199 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 200 | |
1da177e4 LT |
201 | _major = major; |
202 | r = register_blkdev(_major, _name); | |
51157b4a KU |
203 | if (r < 0) |
204 | goto out_uevent_exit; | |
1da177e4 LT |
205 | |
206 | if (!_major) | |
207 | _major = r; | |
208 | ||
209 | return 0; | |
51157b4a KU |
210 | |
211 | out_uevent_exit: | |
212 | dm_uevent_exit(); | |
8fbf26ad KU |
213 | out_free_rq_bio_info_cache: |
214 | kmem_cache_destroy(_rq_bio_info_cache); | |
215 | out_free_rq_tio_cache: | |
216 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
217 | out_free_tio_cache: |
218 | kmem_cache_destroy(_tio_cache); | |
219 | out_free_io_cache: | |
220 | kmem_cache_destroy(_io_cache); | |
221 | ||
222 | return r; | |
1da177e4 LT |
223 | } |
224 | ||
225 | static void local_exit(void) | |
226 | { | |
8fbf26ad KU |
227 | kmem_cache_destroy(_rq_bio_info_cache); |
228 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
229 | kmem_cache_destroy(_tio_cache); |
230 | kmem_cache_destroy(_io_cache); | |
00d59405 | 231 | unregister_blkdev(_major, _name); |
51e5b2bd | 232 | dm_uevent_exit(); |
1da177e4 LT |
233 | |
234 | _major = 0; | |
235 | ||
236 | DMINFO("cleaned up"); | |
237 | } | |
238 | ||
b9249e55 | 239 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
240 | local_init, |
241 | dm_target_init, | |
242 | dm_linear_init, | |
243 | dm_stripe_init, | |
945fa4d2 | 244 | dm_kcopyd_init, |
1da177e4 LT |
245 | dm_interface_init, |
246 | }; | |
247 | ||
b9249e55 | 248 | static void (*_exits[])(void) = { |
1da177e4 LT |
249 | local_exit, |
250 | dm_target_exit, | |
251 | dm_linear_exit, | |
252 | dm_stripe_exit, | |
945fa4d2 | 253 | dm_kcopyd_exit, |
1da177e4 LT |
254 | dm_interface_exit, |
255 | }; | |
256 | ||
257 | static int __init dm_init(void) | |
258 | { | |
259 | const int count = ARRAY_SIZE(_inits); | |
260 | ||
261 | int r, i; | |
262 | ||
263 | for (i = 0; i < count; i++) { | |
264 | r = _inits[i](); | |
265 | if (r) | |
266 | goto bad; | |
267 | } | |
268 | ||
269 | return 0; | |
270 | ||
271 | bad: | |
272 | while (i--) | |
273 | _exits[i](); | |
274 | ||
275 | return r; | |
276 | } | |
277 | ||
278 | static void __exit dm_exit(void) | |
279 | { | |
280 | int i = ARRAY_SIZE(_exits); | |
281 | ||
282 | while (i--) | |
283 | _exits[i](); | |
284 | } | |
285 | ||
286 | /* | |
287 | * Block device functions | |
288 | */ | |
fe5f9f2c | 289 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
290 | { |
291 | struct mapped_device *md; | |
292 | ||
fba9f90e JM |
293 | spin_lock(&_minor_lock); |
294 | ||
fe5f9f2c | 295 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
296 | if (!md) |
297 | goto out; | |
298 | ||
5c6bd75d AK |
299 | if (test_bit(DMF_FREEING, &md->flags) || |
300 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
301 | md = NULL; |
302 | goto out; | |
303 | } | |
304 | ||
1da177e4 | 305 | dm_get(md); |
5c6bd75d | 306 | atomic_inc(&md->open_count); |
fba9f90e JM |
307 | |
308 | out: | |
309 | spin_unlock(&_minor_lock); | |
310 | ||
311 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
312 | } |
313 | ||
fe5f9f2c | 314 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 315 | { |
fe5f9f2c | 316 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 317 | atomic_dec(&md->open_count); |
1da177e4 LT |
318 | dm_put(md); |
319 | return 0; | |
320 | } | |
321 | ||
5c6bd75d AK |
322 | int dm_open_count(struct mapped_device *md) |
323 | { | |
324 | return atomic_read(&md->open_count); | |
325 | } | |
326 | ||
327 | /* | |
328 | * Guarantees nothing is using the device before it's deleted. | |
329 | */ | |
330 | int dm_lock_for_deletion(struct mapped_device *md) | |
331 | { | |
332 | int r = 0; | |
333 | ||
334 | spin_lock(&_minor_lock); | |
335 | ||
336 | if (dm_open_count(md)) | |
337 | r = -EBUSY; | |
338 | else | |
339 | set_bit(DMF_DELETING, &md->flags); | |
340 | ||
341 | spin_unlock(&_minor_lock); | |
342 | ||
343 | return r; | |
344 | } | |
345 | ||
3ac51e74 DW |
346 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
347 | { | |
348 | struct mapped_device *md = bdev->bd_disk->private_data; | |
349 | ||
350 | return dm_get_geometry(md, geo); | |
351 | } | |
352 | ||
fe5f9f2c | 353 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
354 | unsigned int cmd, unsigned long arg) |
355 | { | |
fe5f9f2c AV |
356 | struct mapped_device *md = bdev->bd_disk->private_data; |
357 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
358 | struct dm_target *tgt; |
359 | int r = -ENOTTY; | |
360 | ||
aa129a22 MB |
361 | if (!map || !dm_table_get_size(map)) |
362 | goto out; | |
363 | ||
364 | /* We only support devices that have a single target */ | |
365 | if (dm_table_get_num_targets(map) != 1) | |
366 | goto out; | |
367 | ||
368 | tgt = dm_table_get_target(map, 0); | |
369 | ||
370 | if (dm_suspended(md)) { | |
371 | r = -EAGAIN; | |
372 | goto out; | |
373 | } | |
374 | ||
375 | if (tgt->type->ioctl) | |
647b3d00 | 376 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
377 | |
378 | out: | |
379 | dm_table_put(map); | |
380 | ||
aa129a22 MB |
381 | return r; |
382 | } | |
383 | ||
028867ac | 384 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
385 | { |
386 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
387 | } | |
388 | ||
028867ac | 389 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
390 | { |
391 | mempool_free(io, md->io_pool); | |
392 | } | |
393 | ||
028867ac | 394 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
395 | { |
396 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
397 | } | |
398 | ||
028867ac | 399 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
400 | { |
401 | mempool_free(tio, md->tio_pool); | |
402 | } | |
403 | ||
3eaf840e JNN |
404 | static void start_io_acct(struct dm_io *io) |
405 | { | |
406 | struct mapped_device *md = io->md; | |
c9959059 | 407 | int cpu; |
3eaf840e JNN |
408 | |
409 | io->start_time = jiffies; | |
410 | ||
074a7aca TH |
411 | cpu = part_stat_lock(); |
412 | part_round_stats(cpu, &dm_disk(md)->part0); | |
413 | part_stat_unlock(); | |
414 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
415 | } |
416 | ||
d221d2e7 | 417 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
418 | { |
419 | struct mapped_device *md = io->md; | |
420 | struct bio *bio = io->bio; | |
421 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 422 | int pending, cpu; |
3eaf840e JNN |
423 | int rw = bio_data_dir(bio); |
424 | ||
074a7aca TH |
425 | cpu = part_stat_lock(); |
426 | part_round_stats(cpu, &dm_disk(md)->part0); | |
427 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
428 | part_stat_unlock(); | |
3eaf840e | 429 | |
af7e466a MP |
430 | /* |
431 | * After this is decremented the bio must not be touched if it is | |
432 | * a barrier. | |
433 | */ | |
074a7aca TH |
434 | dm_disk(md)->part0.in_flight = pending = |
435 | atomic_dec_return(&md->pending); | |
3eaf840e | 436 | |
d221d2e7 MP |
437 | /* nudge anyone waiting on suspend queue */ |
438 | if (!pending) | |
439 | wake_up(&md->wait); | |
3eaf840e JNN |
440 | } |
441 | ||
1da177e4 LT |
442 | /* |
443 | * Add the bio to the list of deferred io. | |
444 | */ | |
92c63902 | 445 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 446 | { |
2ca3310e | 447 | down_write(&md->io_lock); |
1da177e4 | 448 | |
022c2611 | 449 | spin_lock_irq(&md->deferred_lock); |
1da177e4 | 450 | bio_list_add(&md->deferred, bio); |
022c2611 | 451 | spin_unlock_irq(&md->deferred_lock); |
1da177e4 | 452 | |
92c63902 MP |
453 | if (!test_and_set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) |
454 | queue_work(md->wq, &md->work); | |
455 | ||
2ca3310e | 456 | up_write(&md->io_lock); |
1da177e4 LT |
457 | } |
458 | ||
459 | /* | |
460 | * Everyone (including functions in this file), should use this | |
461 | * function to access the md->map field, and make sure they call | |
462 | * dm_table_put() when finished. | |
463 | */ | |
464 | struct dm_table *dm_get_table(struct mapped_device *md) | |
465 | { | |
466 | struct dm_table *t; | |
467 | ||
468 | read_lock(&md->map_lock); | |
469 | t = md->map; | |
470 | if (t) | |
471 | dm_table_get(t); | |
472 | read_unlock(&md->map_lock); | |
473 | ||
474 | return t; | |
475 | } | |
476 | ||
3ac51e74 DW |
477 | /* |
478 | * Get the geometry associated with a dm device | |
479 | */ | |
480 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
481 | { | |
482 | *geo = md->geometry; | |
483 | ||
484 | return 0; | |
485 | } | |
486 | ||
487 | /* | |
488 | * Set the geometry of a device. | |
489 | */ | |
490 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
491 | { | |
492 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
493 | ||
494 | if (geo->start > sz) { | |
495 | DMWARN("Start sector is beyond the geometry limits."); | |
496 | return -EINVAL; | |
497 | } | |
498 | ||
499 | md->geometry = *geo; | |
500 | ||
501 | return 0; | |
502 | } | |
503 | ||
1da177e4 LT |
504 | /*----------------------------------------------------------------- |
505 | * CRUD START: | |
506 | * A more elegant soln is in the works that uses the queue | |
507 | * merge fn, unfortunately there are a couple of changes to | |
508 | * the block layer that I want to make for this. So in the | |
509 | * interests of getting something for people to use I give | |
510 | * you this clearly demarcated crap. | |
511 | *---------------------------------------------------------------*/ | |
512 | ||
2e93ccc1 KU |
513 | static int __noflush_suspending(struct mapped_device *md) |
514 | { | |
515 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
516 | } | |
517 | ||
1da177e4 LT |
518 | /* |
519 | * Decrements the number of outstanding ios that a bio has been | |
520 | * cloned into, completing the original io if necc. | |
521 | */ | |
858119e1 | 522 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 523 | { |
2e93ccc1 | 524 | unsigned long flags; |
b35f8caa MB |
525 | int io_error; |
526 | struct bio *bio; | |
527 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
528 | |
529 | /* Push-back supersedes any I/O errors */ | |
b35f8caa | 530 | if (error && !(io->error > 0 && __noflush_suspending(md))) |
1da177e4 LT |
531 | io->error = error; |
532 | ||
533 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
534 | if (io->error == DM_ENDIO_REQUEUE) { |
535 | /* | |
536 | * Target requested pushing back the I/O. | |
2e93ccc1 | 537 | */ |
022c2611 | 538 | spin_lock_irqsave(&md->deferred_lock, flags); |
b35f8caa | 539 | if (__noflush_suspending(md)) |
af7e466a | 540 | bio_list_add_head(&md->deferred, io->bio); |
2e93ccc1 KU |
541 | else |
542 | /* noflush suspend was interrupted. */ | |
543 | io->error = -EIO; | |
022c2611 | 544 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
545 | } |
546 | ||
b35f8caa MB |
547 | io_error = io->error; |
548 | bio = io->bio; | |
2e93ccc1 | 549 | |
af7e466a MP |
550 | if (bio_barrier(bio)) { |
551 | /* | |
552 | * There can be just one barrier request so we use | |
553 | * a per-device variable for error reporting. | |
554 | * Note that you can't touch the bio after end_io_acct | |
555 | */ | |
556 | md->barrier_error = io_error; | |
557 | end_io_acct(io); | |
558 | } else { | |
559 | end_io_acct(io); | |
b35f8caa | 560 | |
af7e466a MP |
561 | if (io_error != DM_ENDIO_REQUEUE) { |
562 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 563 | |
af7e466a MP |
564 | bio_endio(bio, io_error); |
565 | } | |
b35f8caa | 566 | } |
af7e466a MP |
567 | |
568 | free_io(md, io); | |
1da177e4 LT |
569 | } |
570 | } | |
571 | ||
6712ecf8 | 572 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
573 | { |
574 | int r = 0; | |
028867ac | 575 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 576 | struct dm_io *io = tio->io; |
9faf400f | 577 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
578 | dm_endio_fn endio = tio->ti->type->end_io; |
579 | ||
1da177e4 LT |
580 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
581 | error = -EIO; | |
582 | ||
583 | if (endio) { | |
584 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
585 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
586 | /* | |
587 | * error and requeue request are handled | |
588 | * in dec_pending(). | |
589 | */ | |
1da177e4 | 590 | error = r; |
45cbcd79 KU |
591 | else if (r == DM_ENDIO_INCOMPLETE) |
592 | /* The target will handle the io */ | |
6712ecf8 | 593 | return; |
45cbcd79 KU |
594 | else if (r) { |
595 | DMWARN("unimplemented target endio return value: %d", r); | |
596 | BUG(); | |
597 | } | |
1da177e4 LT |
598 | } |
599 | ||
9faf400f SB |
600 | /* |
601 | * Store md for cleanup instead of tio which is about to get freed. | |
602 | */ | |
603 | bio->bi_private = md->bs; | |
604 | ||
9faf400f | 605 | free_tio(md, tio); |
b35f8caa MB |
606 | bio_put(bio); |
607 | dec_pending(io, error); | |
1da177e4 LT |
608 | } |
609 | ||
610 | static sector_t max_io_len(struct mapped_device *md, | |
611 | sector_t sector, struct dm_target *ti) | |
612 | { | |
613 | sector_t offset = sector - ti->begin; | |
614 | sector_t len = ti->len - offset; | |
615 | ||
616 | /* | |
617 | * Does the target need to split even further ? | |
618 | */ | |
619 | if (ti->split_io) { | |
620 | sector_t boundary; | |
621 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
622 | - offset; | |
623 | if (len > boundary) | |
624 | len = boundary; | |
625 | } | |
626 | ||
627 | return len; | |
628 | } | |
629 | ||
630 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 631 | struct dm_target_io *tio) |
1da177e4 LT |
632 | { |
633 | int r; | |
2056a782 | 634 | sector_t sector; |
9faf400f | 635 | struct mapped_device *md; |
1da177e4 LT |
636 | |
637 | /* | |
638 | * Sanity checks. | |
639 | */ | |
640 | BUG_ON(!clone->bi_size); | |
641 | ||
642 | clone->bi_end_io = clone_endio; | |
643 | clone->bi_private = tio; | |
644 | ||
645 | /* | |
646 | * Map the clone. If r == 0 we don't need to do | |
647 | * anything, the target has assumed ownership of | |
648 | * this io. | |
649 | */ | |
650 | atomic_inc(&tio->io->io_count); | |
2056a782 | 651 | sector = clone->bi_sector; |
1da177e4 | 652 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 653 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 654 | /* the bio has been remapped so dispatch it */ |
2056a782 | 655 | |
5f3ea37c | 656 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
22a7c31a | 657 | tio->io->bio->bi_bdev->bd_dev, sector); |
2056a782 | 658 | |
1da177e4 | 659 | generic_make_request(clone); |
2e93ccc1 KU |
660 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
661 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
662 | md = tio->io->md; |
663 | dec_pending(tio->io, r); | |
664 | /* | |
665 | * Store bio_set for cleanup. | |
666 | */ | |
667 | clone->bi_private = md->bs; | |
1da177e4 | 668 | bio_put(clone); |
9faf400f | 669 | free_tio(md, tio); |
45cbcd79 KU |
670 | } else if (r) { |
671 | DMWARN("unimplemented target map return value: %d", r); | |
672 | BUG(); | |
1da177e4 LT |
673 | } |
674 | } | |
675 | ||
676 | struct clone_info { | |
677 | struct mapped_device *md; | |
678 | struct dm_table *map; | |
679 | struct bio *bio; | |
680 | struct dm_io *io; | |
681 | sector_t sector; | |
682 | sector_t sector_count; | |
683 | unsigned short idx; | |
684 | }; | |
685 | ||
3676347a PO |
686 | static void dm_bio_destructor(struct bio *bio) |
687 | { | |
9faf400f SB |
688 | struct bio_set *bs = bio->bi_private; |
689 | ||
690 | bio_free(bio, bs); | |
3676347a PO |
691 | } |
692 | ||
1da177e4 LT |
693 | /* |
694 | * Creates a little bio that is just does part of a bvec. | |
695 | */ | |
696 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
697 | unsigned short idx, unsigned int offset, | |
9faf400f | 698 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
699 | { |
700 | struct bio *clone; | |
701 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
702 | ||
9faf400f | 703 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 704 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
705 | *clone->bi_io_vec = *bv; |
706 | ||
707 | clone->bi_sector = sector; | |
708 | clone->bi_bdev = bio->bi_bdev; | |
af7e466a | 709 | clone->bi_rw = bio->bi_rw & ~(1 << BIO_RW_BARRIER); |
1da177e4 LT |
710 | clone->bi_vcnt = 1; |
711 | clone->bi_size = to_bytes(len); | |
712 | clone->bi_io_vec->bv_offset = offset; | |
713 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 714 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 715 | |
9c47008d MP |
716 | if (bio_integrity(bio)) { |
717 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
718 | bio_integrity_trim(clone, | |
719 | bio_sector_offset(bio, idx, offset), len); | |
720 | } | |
721 | ||
1da177e4 LT |
722 | return clone; |
723 | } | |
724 | ||
725 | /* | |
726 | * Creates a bio that consists of range of complete bvecs. | |
727 | */ | |
728 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
729 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 730 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
731 | { |
732 | struct bio *clone; | |
733 | ||
9faf400f SB |
734 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
735 | __bio_clone(clone, bio); | |
af7e466a | 736 | clone->bi_rw &= ~(1 << BIO_RW_BARRIER); |
9faf400f | 737 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
738 | clone->bi_sector = sector; |
739 | clone->bi_idx = idx; | |
740 | clone->bi_vcnt = idx + bv_count; | |
741 | clone->bi_size = to_bytes(len); | |
742 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
743 | ||
9c47008d MP |
744 | if (bio_integrity(bio)) { |
745 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
746 | ||
747 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
748 | bio_integrity_trim(clone, | |
749 | bio_sector_offset(bio, idx, 0), len); | |
750 | } | |
751 | ||
1da177e4 LT |
752 | return clone; |
753 | } | |
754 | ||
512875bd | 755 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
756 | { |
757 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
758 | struct dm_target *ti; |
759 | sector_t len = 0, max; | |
028867ac | 760 | struct dm_target_io *tio; |
1da177e4 | 761 | |
512875bd JN |
762 | ti = dm_table_find_target(ci->map, ci->sector); |
763 | if (!dm_target_is_valid(ti)) | |
764 | return -EIO; | |
765 | ||
766 | max = max_io_len(ci->md, ci->sector, ti); | |
767 | ||
1da177e4 LT |
768 | /* |
769 | * Allocate a target io object. | |
770 | */ | |
771 | tio = alloc_tio(ci->md); | |
772 | tio->io = ci->io; | |
773 | tio->ti = ti; | |
774 | memset(&tio->info, 0, sizeof(tio->info)); | |
775 | ||
776 | if (ci->sector_count <= max) { | |
777 | /* | |
778 | * Optimise for the simple case where we can do all of | |
779 | * the remaining io with a single clone. | |
780 | */ | |
781 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
782 | bio->bi_vcnt - ci->idx, ci->sector_count, |
783 | ci->md->bs); | |
1da177e4 LT |
784 | __map_bio(ti, clone, tio); |
785 | ci->sector_count = 0; | |
786 | ||
787 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
788 | /* | |
789 | * There are some bvecs that don't span targets. | |
790 | * Do as many of these as possible. | |
791 | */ | |
792 | int i; | |
793 | sector_t remaining = max; | |
794 | sector_t bv_len; | |
795 | ||
796 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
797 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
798 | ||
799 | if (bv_len > remaining) | |
800 | break; | |
801 | ||
802 | remaining -= bv_len; | |
803 | len += bv_len; | |
804 | } | |
805 | ||
9faf400f SB |
806 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
807 | ci->md->bs); | |
1da177e4 LT |
808 | __map_bio(ti, clone, tio); |
809 | ||
810 | ci->sector += len; | |
811 | ci->sector_count -= len; | |
812 | ci->idx = i; | |
813 | ||
814 | } else { | |
815 | /* | |
d2044a94 | 816 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
817 | */ |
818 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
819 | sector_t remaining = to_sector(bv->bv_len); |
820 | unsigned int offset = 0; | |
1da177e4 | 821 | |
d2044a94 AK |
822 | do { |
823 | if (offset) { | |
824 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
825 | if (!dm_target_is_valid(ti)) |
826 | return -EIO; | |
827 | ||
d2044a94 | 828 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 829 | |
d2044a94 AK |
830 | tio = alloc_tio(ci->md); |
831 | tio->io = ci->io; | |
832 | tio->ti = ti; | |
833 | memset(&tio->info, 0, sizeof(tio->info)); | |
834 | } | |
835 | ||
836 | len = min(remaining, max); | |
837 | ||
838 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
839 | bv->bv_offset + offset, len, |
840 | ci->md->bs); | |
d2044a94 AK |
841 | |
842 | __map_bio(ti, clone, tio); | |
843 | ||
844 | ci->sector += len; | |
845 | ci->sector_count -= len; | |
846 | offset += to_bytes(len); | |
847 | } while (remaining -= len); | |
1da177e4 | 848 | |
1da177e4 LT |
849 | ci->idx++; |
850 | } | |
512875bd JN |
851 | |
852 | return 0; | |
1da177e4 LT |
853 | } |
854 | ||
855 | /* | |
8a53c28d | 856 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 857 | */ |
f0b9a450 | 858 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
859 | { |
860 | struct clone_info ci; | |
512875bd | 861 | int error = 0; |
1da177e4 LT |
862 | |
863 | ci.map = dm_get_table(md); | |
f0b9a450 | 864 | if (unlikely(!ci.map)) { |
af7e466a MP |
865 | if (!bio_barrier(bio)) |
866 | bio_io_error(bio); | |
867 | else | |
868 | md->barrier_error = -EIO; | |
f0b9a450 MP |
869 | return; |
870 | } | |
692d0eb9 | 871 | |
1da177e4 LT |
872 | ci.md = md; |
873 | ci.bio = bio; | |
874 | ci.io = alloc_io(md); | |
875 | ci.io->error = 0; | |
876 | atomic_set(&ci.io->io_count, 1); | |
877 | ci.io->bio = bio; | |
878 | ci.io->md = md; | |
879 | ci.sector = bio->bi_sector; | |
880 | ci.sector_count = bio_sectors(bio); | |
881 | ci.idx = bio->bi_idx; | |
882 | ||
3eaf840e | 883 | start_io_acct(ci.io); |
512875bd JN |
884 | while (ci.sector_count && !error) |
885 | error = __clone_and_map(&ci); | |
1da177e4 LT |
886 | |
887 | /* drop the extra reference count */ | |
512875bd | 888 | dec_pending(ci.io, error); |
1da177e4 LT |
889 | dm_table_put(ci.map); |
890 | } | |
891 | /*----------------------------------------------------------------- | |
892 | * CRUD END | |
893 | *---------------------------------------------------------------*/ | |
894 | ||
f6fccb12 MB |
895 | static int dm_merge_bvec(struct request_queue *q, |
896 | struct bvec_merge_data *bvm, | |
897 | struct bio_vec *biovec) | |
898 | { | |
899 | struct mapped_device *md = q->queuedata; | |
900 | struct dm_table *map = dm_get_table(md); | |
901 | struct dm_target *ti; | |
902 | sector_t max_sectors; | |
5037108a | 903 | int max_size = 0; |
f6fccb12 MB |
904 | |
905 | if (unlikely(!map)) | |
5037108a | 906 | goto out; |
f6fccb12 MB |
907 | |
908 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
909 | if (!dm_target_is_valid(ti)) |
910 | goto out_table; | |
f6fccb12 MB |
911 | |
912 | /* | |
913 | * Find maximum amount of I/O that won't need splitting | |
914 | */ | |
915 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
916 | (sector_t) BIO_MAX_SECTORS); | |
917 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
918 | if (max_size < 0) | |
919 | max_size = 0; | |
920 | ||
921 | /* | |
922 | * merge_bvec_fn() returns number of bytes | |
923 | * it can accept at this offset | |
924 | * max is precomputed maximal io size | |
925 | */ | |
926 | if (max_size && ti->type->merge) | |
927 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
928 | ||
b01cd5ac | 929 | out_table: |
5037108a MP |
930 | dm_table_put(map); |
931 | ||
932 | out: | |
f6fccb12 MB |
933 | /* |
934 | * Always allow an entire first page | |
935 | */ | |
936 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
937 | max_size = biovec->bv_len; | |
938 | ||
f6fccb12 MB |
939 | return max_size; |
940 | } | |
941 | ||
1da177e4 LT |
942 | /* |
943 | * The request function that just remaps the bio built up by | |
944 | * dm_merge_bvec. | |
945 | */ | |
165125e1 | 946 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 947 | { |
12f03a49 | 948 | int rw = bio_data_dir(bio); |
1da177e4 | 949 | struct mapped_device *md = q->queuedata; |
c9959059 | 950 | int cpu; |
1da177e4 | 951 | |
2ca3310e | 952 | down_read(&md->io_lock); |
1da177e4 | 953 | |
074a7aca TH |
954 | cpu = part_stat_lock(); |
955 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
956 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
957 | part_stat_unlock(); | |
12f03a49 | 958 | |
1da177e4 | 959 | /* |
1eb787ec AK |
960 | * If we're suspended or the thread is processing barriers |
961 | * we have to queue this io for later. | |
1da177e4 | 962 | */ |
af7e466a MP |
963 | if (unlikely(test_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) || |
964 | unlikely(bio_barrier(bio))) { | |
2ca3310e | 965 | up_read(&md->io_lock); |
1da177e4 | 966 | |
54d9a1b4 AK |
967 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) && |
968 | bio_rw(bio) == READA) { | |
969 | bio_io_error(bio); | |
970 | return 0; | |
971 | } | |
1da177e4 | 972 | |
92c63902 | 973 | queue_io(md, bio); |
1da177e4 | 974 | |
92c63902 | 975 | return 0; |
1da177e4 LT |
976 | } |
977 | ||
f0b9a450 | 978 | __split_and_process_bio(md, bio); |
2ca3310e | 979 | up_read(&md->io_lock); |
f0b9a450 | 980 | return 0; |
1da177e4 LT |
981 | } |
982 | ||
165125e1 | 983 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
984 | { |
985 | struct mapped_device *md = q->queuedata; | |
986 | struct dm_table *map = dm_get_table(md); | |
987 | ||
988 | if (map) { | |
989 | dm_table_unplug_all(map); | |
990 | dm_table_put(map); | |
991 | } | |
992 | } | |
993 | ||
994 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
995 | { | |
8a57dfc6 CS |
996 | int r = bdi_bits; |
997 | struct mapped_device *md = congested_data; | |
998 | struct dm_table *map; | |
1da177e4 | 999 | |
1eb787ec | 1000 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
8a57dfc6 CS |
1001 | map = dm_get_table(md); |
1002 | if (map) { | |
1003 | r = dm_table_any_congested(map, bdi_bits); | |
1004 | dm_table_put(map); | |
1005 | } | |
1006 | } | |
1007 | ||
1da177e4 LT |
1008 | return r; |
1009 | } | |
1010 | ||
1011 | /*----------------------------------------------------------------- | |
1012 | * An IDR is used to keep track of allocated minor numbers. | |
1013 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1014 | static DEFINE_IDR(_minor_idr); |
1015 | ||
2b06cfff | 1016 | static void free_minor(int minor) |
1da177e4 | 1017 | { |
f32c10b0 | 1018 | spin_lock(&_minor_lock); |
1da177e4 | 1019 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1020 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1021 | } |
1022 | ||
1023 | /* | |
1024 | * See if the device with a specific minor # is free. | |
1025 | */ | |
cf13ab8e | 1026 | static int specific_minor(int minor) |
1da177e4 LT |
1027 | { |
1028 | int r, m; | |
1029 | ||
1030 | if (minor >= (1 << MINORBITS)) | |
1031 | return -EINVAL; | |
1032 | ||
62f75c2f JM |
1033 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1034 | if (!r) | |
1035 | return -ENOMEM; | |
1036 | ||
f32c10b0 | 1037 | spin_lock(&_minor_lock); |
1da177e4 LT |
1038 | |
1039 | if (idr_find(&_minor_idr, minor)) { | |
1040 | r = -EBUSY; | |
1041 | goto out; | |
1042 | } | |
1043 | ||
ba61fdd1 | 1044 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1045 | if (r) |
1da177e4 | 1046 | goto out; |
1da177e4 LT |
1047 | |
1048 | if (m != minor) { | |
1049 | idr_remove(&_minor_idr, m); | |
1050 | r = -EBUSY; | |
1051 | goto out; | |
1052 | } | |
1053 | ||
1054 | out: | |
f32c10b0 | 1055 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1056 | return r; |
1057 | } | |
1058 | ||
cf13ab8e | 1059 | static int next_free_minor(int *minor) |
1da177e4 | 1060 | { |
2b06cfff | 1061 | int r, m; |
1da177e4 | 1062 | |
1da177e4 | 1063 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1064 | if (!r) |
1065 | return -ENOMEM; | |
1066 | ||
f32c10b0 | 1067 | spin_lock(&_minor_lock); |
1da177e4 | 1068 | |
ba61fdd1 | 1069 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1070 | if (r) |
1da177e4 | 1071 | goto out; |
1da177e4 LT |
1072 | |
1073 | if (m >= (1 << MINORBITS)) { | |
1074 | idr_remove(&_minor_idr, m); | |
1075 | r = -ENOSPC; | |
1076 | goto out; | |
1077 | } | |
1078 | ||
1079 | *minor = m; | |
1080 | ||
1081 | out: | |
f32c10b0 | 1082 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1083 | return r; |
1084 | } | |
1085 | ||
1086 | static struct block_device_operations dm_blk_dops; | |
1087 | ||
53d5914f MP |
1088 | static void dm_wq_work(struct work_struct *work); |
1089 | ||
1da177e4 LT |
1090 | /* |
1091 | * Allocate and initialise a blank device with a given minor. | |
1092 | */ | |
2b06cfff | 1093 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1094 | { |
1095 | int r; | |
cf13ab8e | 1096 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1097 | void *old_md; |
1da177e4 LT |
1098 | |
1099 | if (!md) { | |
1100 | DMWARN("unable to allocate device, out of memory."); | |
1101 | return NULL; | |
1102 | } | |
1103 | ||
10da4f79 | 1104 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1105 | goto bad_module_get; |
10da4f79 | 1106 | |
1da177e4 | 1107 | /* get a minor number for the dev */ |
2b06cfff | 1108 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1109 | r = next_free_minor(&minor); |
2b06cfff | 1110 | else |
cf13ab8e | 1111 | r = specific_minor(minor); |
1da177e4 | 1112 | if (r < 0) |
6ed7ade8 | 1113 | goto bad_minor; |
1da177e4 | 1114 | |
2ca3310e | 1115 | init_rwsem(&md->io_lock); |
e61290a4 | 1116 | mutex_init(&md->suspend_lock); |
022c2611 | 1117 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1118 | rwlock_init(&md->map_lock); |
1119 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1120 | atomic_set(&md->open_count, 0); |
1da177e4 | 1121 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1122 | atomic_set(&md->uevent_seq, 0); |
1123 | INIT_LIST_HEAD(&md->uevent_list); | |
1124 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1125 | |
1126 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1127 | if (!md->queue) | |
6ed7ade8 | 1128 | goto bad_queue; |
1da177e4 LT |
1129 | |
1130 | md->queue->queuedata = md; | |
1131 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1132 | md->queue->backing_dev_info.congested_data = md; | |
1133 | blk_queue_make_request(md->queue, dm_request); | |
99360b4c | 1134 | blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN, NULL); |
daef265f | 1135 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1136 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1137 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1138 | |
93d2341c | 1139 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1140 | if (!md->io_pool) |
6ed7ade8 | 1141 | goto bad_io_pool; |
1da177e4 | 1142 | |
93d2341c | 1143 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1144 | if (!md->tio_pool) |
6ed7ade8 | 1145 | goto bad_tio_pool; |
1da177e4 | 1146 | |
bb799ca0 | 1147 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1148 | if (!md->bs) |
1149 | goto bad_no_bioset; | |
1150 | ||
1da177e4 LT |
1151 | md->disk = alloc_disk(1); |
1152 | if (!md->disk) | |
6ed7ade8 | 1153 | goto bad_disk; |
1da177e4 | 1154 | |
f0b04115 JM |
1155 | atomic_set(&md->pending, 0); |
1156 | init_waitqueue_head(&md->wait); | |
53d5914f | 1157 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1158 | init_waitqueue_head(&md->eventq); |
1159 | ||
1da177e4 LT |
1160 | md->disk->major = _major; |
1161 | md->disk->first_minor = minor; | |
1162 | md->disk->fops = &dm_blk_dops; | |
1163 | md->disk->queue = md->queue; | |
1164 | md->disk->private_data = md; | |
1165 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1166 | add_disk(md->disk); | |
7e51f257 | 1167 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1168 | |
304f3f6a MB |
1169 | md->wq = create_singlethread_workqueue("kdmflush"); |
1170 | if (!md->wq) | |
1171 | goto bad_thread; | |
1172 | ||
ba61fdd1 | 1173 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1174 | spin_lock(&_minor_lock); |
ba61fdd1 | 1175 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1176 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1177 | |
1178 | BUG_ON(old_md != MINOR_ALLOCED); | |
1179 | ||
1da177e4 LT |
1180 | return md; |
1181 | ||
304f3f6a MB |
1182 | bad_thread: |
1183 | put_disk(md->disk); | |
6ed7ade8 | 1184 | bad_disk: |
9faf400f | 1185 | bioset_free(md->bs); |
6ed7ade8 | 1186 | bad_no_bioset: |
1da177e4 | 1187 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1188 | bad_tio_pool: |
1da177e4 | 1189 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1190 | bad_io_pool: |
1312f40e | 1191 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1192 | bad_queue: |
1da177e4 | 1193 | free_minor(minor); |
6ed7ade8 | 1194 | bad_minor: |
10da4f79 | 1195 | module_put(THIS_MODULE); |
6ed7ade8 | 1196 | bad_module_get: |
1da177e4 LT |
1197 | kfree(md); |
1198 | return NULL; | |
1199 | } | |
1200 | ||
ae9da83f JN |
1201 | static void unlock_fs(struct mapped_device *md); |
1202 | ||
1da177e4 LT |
1203 | static void free_dev(struct mapped_device *md) |
1204 | { | |
f331c029 | 1205 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1206 | |
d9dde59b | 1207 | if (md->suspended_bdev) { |
ae9da83f | 1208 | unlock_fs(md); |
d9dde59b JN |
1209 | bdput(md->suspended_bdev); |
1210 | } | |
304f3f6a | 1211 | destroy_workqueue(md->wq); |
1da177e4 LT |
1212 | mempool_destroy(md->tio_pool); |
1213 | mempool_destroy(md->io_pool); | |
9faf400f | 1214 | bioset_free(md->bs); |
9c47008d | 1215 | blk_integrity_unregister(md->disk); |
1da177e4 | 1216 | del_gendisk(md->disk); |
63d94e48 | 1217 | free_minor(minor); |
fba9f90e JM |
1218 | |
1219 | spin_lock(&_minor_lock); | |
1220 | md->disk->private_data = NULL; | |
1221 | spin_unlock(&_minor_lock); | |
1222 | ||
1da177e4 | 1223 | put_disk(md->disk); |
1312f40e | 1224 | blk_cleanup_queue(md->queue); |
10da4f79 | 1225 | module_put(THIS_MODULE); |
1da177e4 LT |
1226 | kfree(md); |
1227 | } | |
1228 | ||
1229 | /* | |
1230 | * Bind a table to the device. | |
1231 | */ | |
1232 | static void event_callback(void *context) | |
1233 | { | |
7a8c3d3b MA |
1234 | unsigned long flags; |
1235 | LIST_HEAD(uevents); | |
1da177e4 LT |
1236 | struct mapped_device *md = (struct mapped_device *) context; |
1237 | ||
7a8c3d3b MA |
1238 | spin_lock_irqsave(&md->uevent_lock, flags); |
1239 | list_splice_init(&md->uevent_list, &uevents); | |
1240 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1241 | ||
ed9e1982 | 1242 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1243 | |
1da177e4 LT |
1244 | atomic_inc(&md->event_nr); |
1245 | wake_up(&md->eventq); | |
1246 | } | |
1247 | ||
4e90188b | 1248 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1249 | { |
4e90188b | 1250 | set_capacity(md->disk, size); |
1da177e4 | 1251 | |
1b1dcc1b | 1252 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1253 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1254 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1255 | } |
1256 | ||
1257 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1258 | { | |
165125e1 | 1259 | struct request_queue *q = md->queue; |
1da177e4 LT |
1260 | sector_t size; |
1261 | ||
1262 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1263 | |
1264 | /* | |
1265 | * Wipe any geometry if the size of the table changed. | |
1266 | */ | |
1267 | if (size != get_capacity(md->disk)) | |
1268 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1269 | ||
bfa152fa JN |
1270 | if (md->suspended_bdev) |
1271 | __set_size(md, size); | |
d5816876 MP |
1272 | |
1273 | if (!size) { | |
1274 | dm_table_destroy(t); | |
1da177e4 | 1275 | return 0; |
d5816876 | 1276 | } |
1da177e4 | 1277 | |
2ca3310e AK |
1278 | dm_table_event_callback(t, event_callback, md); |
1279 | ||
1da177e4 LT |
1280 | write_lock(&md->map_lock); |
1281 | md->map = t; | |
2ca3310e | 1282 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1283 | write_unlock(&md->map_lock); |
1284 | ||
1da177e4 LT |
1285 | return 0; |
1286 | } | |
1287 | ||
1288 | static void __unbind(struct mapped_device *md) | |
1289 | { | |
1290 | struct dm_table *map = md->map; | |
1291 | ||
1292 | if (!map) | |
1293 | return; | |
1294 | ||
1295 | dm_table_event_callback(map, NULL, NULL); | |
1296 | write_lock(&md->map_lock); | |
1297 | md->map = NULL; | |
1298 | write_unlock(&md->map_lock); | |
d5816876 | 1299 | dm_table_destroy(map); |
1da177e4 LT |
1300 | } |
1301 | ||
1302 | /* | |
1303 | * Constructor for a new device. | |
1304 | */ | |
2b06cfff | 1305 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1306 | { |
1307 | struct mapped_device *md; | |
1308 | ||
2b06cfff | 1309 | md = alloc_dev(minor); |
1da177e4 LT |
1310 | if (!md) |
1311 | return -ENXIO; | |
1312 | ||
784aae73 MB |
1313 | dm_sysfs_init(md); |
1314 | ||
1da177e4 LT |
1315 | *result = md; |
1316 | return 0; | |
1317 | } | |
1318 | ||
637842cf | 1319 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1320 | { |
1321 | struct mapped_device *md; | |
1da177e4 LT |
1322 | unsigned minor = MINOR(dev); |
1323 | ||
1324 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1325 | return NULL; | |
1326 | ||
f32c10b0 | 1327 | spin_lock(&_minor_lock); |
1da177e4 LT |
1328 | |
1329 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1330 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1331 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1332 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1333 | md = NULL; |
fba9f90e JM |
1334 | goto out; |
1335 | } | |
1da177e4 | 1336 | |
fba9f90e | 1337 | out: |
f32c10b0 | 1338 | spin_unlock(&_minor_lock); |
1da177e4 | 1339 | |
637842cf DT |
1340 | return md; |
1341 | } | |
1342 | ||
d229a958 DT |
1343 | struct mapped_device *dm_get_md(dev_t dev) |
1344 | { | |
1345 | struct mapped_device *md = dm_find_md(dev); | |
1346 | ||
1347 | if (md) | |
1348 | dm_get(md); | |
1349 | ||
1350 | return md; | |
1351 | } | |
1352 | ||
9ade92a9 | 1353 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1354 | { |
9ade92a9 | 1355 | return md->interface_ptr; |
1da177e4 LT |
1356 | } |
1357 | ||
1358 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1359 | { | |
1360 | md->interface_ptr = ptr; | |
1361 | } | |
1362 | ||
1363 | void dm_get(struct mapped_device *md) | |
1364 | { | |
1365 | atomic_inc(&md->holders); | |
1366 | } | |
1367 | ||
72d94861 AK |
1368 | const char *dm_device_name(struct mapped_device *md) |
1369 | { | |
1370 | return md->name; | |
1371 | } | |
1372 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1373 | ||
1da177e4 LT |
1374 | void dm_put(struct mapped_device *md) |
1375 | { | |
1134e5ae | 1376 | struct dm_table *map; |
1da177e4 | 1377 | |
fba9f90e JM |
1378 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1379 | ||
f32c10b0 | 1380 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1381 | map = dm_get_table(md); |
f331c029 TH |
1382 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1383 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1384 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1385 | spin_unlock(&_minor_lock); |
cf222b37 | 1386 | if (!dm_suspended(md)) { |
1da177e4 LT |
1387 | dm_table_presuspend_targets(map); |
1388 | dm_table_postsuspend_targets(map); | |
1389 | } | |
784aae73 | 1390 | dm_sysfs_exit(md); |
1134e5ae | 1391 | dm_table_put(map); |
a1b51e98 | 1392 | __unbind(md); |
1da177e4 LT |
1393 | free_dev(md); |
1394 | } | |
1da177e4 | 1395 | } |
79eb885c | 1396 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1397 | |
401600df | 1398 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
1399 | { |
1400 | int r = 0; | |
b44ebeb0 MP |
1401 | DECLARE_WAITQUEUE(wait, current); |
1402 | ||
1403 | dm_unplug_all(md->queue); | |
1404 | ||
1405 | add_wait_queue(&md->wait, &wait); | |
46125c1c MB |
1406 | |
1407 | while (1) { | |
401600df | 1408 | set_current_state(interruptible); |
46125c1c MB |
1409 | |
1410 | smp_mb(); | |
1411 | if (!atomic_read(&md->pending)) | |
1412 | break; | |
1413 | ||
401600df MP |
1414 | if (interruptible == TASK_INTERRUPTIBLE && |
1415 | signal_pending(current)) { | |
46125c1c MB |
1416 | r = -EINTR; |
1417 | break; | |
1418 | } | |
1419 | ||
1420 | io_schedule(); | |
1421 | } | |
1422 | set_current_state(TASK_RUNNING); | |
1423 | ||
b44ebeb0 MP |
1424 | remove_wait_queue(&md->wait, &wait); |
1425 | ||
46125c1c MB |
1426 | return r; |
1427 | } | |
1428 | ||
af7e466a MP |
1429 | static int dm_flush(struct mapped_device *md) |
1430 | { | |
1431 | dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE); | |
1432 | return 0; | |
1433 | } | |
1434 | ||
1435 | static void process_barrier(struct mapped_device *md, struct bio *bio) | |
1436 | { | |
1437 | int error = dm_flush(md); | |
1438 | ||
1439 | if (unlikely(error)) { | |
1440 | bio_endio(bio, error); | |
1441 | return; | |
1442 | } | |
1443 | if (bio_empty_barrier(bio)) { | |
1444 | bio_endio(bio, 0); | |
1445 | return; | |
1446 | } | |
1447 | ||
1448 | __split_and_process_bio(md, bio); | |
1449 | ||
1450 | error = dm_flush(md); | |
1451 | ||
1452 | if (!error && md->barrier_error) | |
1453 | error = md->barrier_error; | |
1454 | ||
1455 | if (md->barrier_error != DM_ENDIO_REQUEUE) | |
1456 | bio_endio(bio, error); | |
1457 | } | |
1458 | ||
1da177e4 LT |
1459 | /* |
1460 | * Process the deferred bios | |
1461 | */ | |
ef208587 | 1462 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 1463 | { |
ef208587 MP |
1464 | struct mapped_device *md = container_of(work, struct mapped_device, |
1465 | work); | |
6d6f10df | 1466 | struct bio *c; |
1da177e4 | 1467 | |
ef208587 MP |
1468 | down_write(&md->io_lock); |
1469 | ||
3b00b203 | 1470 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
1471 | spin_lock_irq(&md->deferred_lock); |
1472 | c = bio_list_pop(&md->deferred); | |
1473 | spin_unlock_irq(&md->deferred_lock); | |
1474 | ||
1475 | if (!c) { | |
1eb787ec | 1476 | clear_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); |
df12ee99 AK |
1477 | break; |
1478 | } | |
022c2611 | 1479 | |
3b00b203 MP |
1480 | up_write(&md->io_lock); |
1481 | ||
af7e466a MP |
1482 | if (bio_barrier(c)) |
1483 | process_barrier(md, c); | |
1484 | else | |
1485 | __split_and_process_bio(md, c); | |
3b00b203 MP |
1486 | |
1487 | down_write(&md->io_lock); | |
022c2611 | 1488 | } |
73d410c0 | 1489 | |
ef208587 | 1490 | up_write(&md->io_lock); |
1da177e4 LT |
1491 | } |
1492 | ||
9a1fb464 | 1493 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 1494 | { |
3b00b203 MP |
1495 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1496 | smp_mb__after_clear_bit(); | |
53d5914f | 1497 | queue_work(md->wq, &md->work); |
304f3f6a MB |
1498 | } |
1499 | ||
1da177e4 LT |
1500 | /* |
1501 | * Swap in a new table (destroying old one). | |
1502 | */ | |
1503 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1504 | { | |
93c534ae | 1505 | int r = -EINVAL; |
1da177e4 | 1506 | |
e61290a4 | 1507 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1508 | |
1509 | /* device must be suspended */ | |
cf222b37 | 1510 | if (!dm_suspended(md)) |
93c534ae | 1511 | goto out; |
1da177e4 | 1512 | |
bfa152fa JN |
1513 | /* without bdev, the device size cannot be changed */ |
1514 | if (!md->suspended_bdev) | |
1515 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1516 | goto out; | |
1517 | ||
1da177e4 LT |
1518 | __unbind(md); |
1519 | r = __bind(md, table); | |
1da177e4 | 1520 | |
93c534ae | 1521 | out: |
e61290a4 | 1522 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1523 | return r; |
1da177e4 LT |
1524 | } |
1525 | ||
1526 | /* | |
1527 | * Functions to lock and unlock any filesystem running on the | |
1528 | * device. | |
1529 | */ | |
2ca3310e | 1530 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1531 | { |
e39e2e95 | 1532 | int r; |
1da177e4 LT |
1533 | |
1534 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1535 | |
e39e2e95 | 1536 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1537 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1538 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1539 | md->frozen_sb = NULL; |
1540 | return r; | |
dfbe03f6 AK |
1541 | } |
1542 | ||
aa8d7c2f AK |
1543 | set_bit(DMF_FROZEN, &md->flags); |
1544 | ||
1da177e4 | 1545 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1546 | * to go away while it is locked. |
1da177e4 LT |
1547 | */ |
1548 | return 0; | |
1549 | } | |
1550 | ||
2ca3310e | 1551 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1552 | { |
aa8d7c2f AK |
1553 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1554 | return; | |
1555 | ||
e39e2e95 | 1556 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1557 | md->frozen_sb = NULL; |
aa8d7c2f | 1558 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1559 | } |
1560 | ||
1561 | /* | |
1562 | * We need to be able to change a mapping table under a mounted | |
1563 | * filesystem. For example we might want to move some data in | |
1564 | * the background. Before the table can be swapped with | |
1565 | * dm_bind_table, dm_suspend must be called to flush any in | |
1566 | * flight bios and ensure that any further io gets deferred. | |
1567 | */ | |
a3d77d35 | 1568 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1569 | { |
2ca3310e | 1570 | struct dm_table *map = NULL; |
46125c1c | 1571 | int r = 0; |
a3d77d35 | 1572 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1573 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1574 | |
e61290a4 | 1575 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1576 | |
73d410c0 MB |
1577 | if (dm_suspended(md)) { |
1578 | r = -EINVAL; | |
d287483d | 1579 | goto out_unlock; |
73d410c0 | 1580 | } |
1da177e4 LT |
1581 | |
1582 | map = dm_get_table(md); | |
1da177e4 | 1583 | |
2e93ccc1 KU |
1584 | /* |
1585 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1586 | * This flag is cleared before dm_suspend returns. | |
1587 | */ | |
1588 | if (noflush) | |
1589 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1590 | ||
cf222b37 AK |
1591 | /* This does not get reverted if there's an error later. */ |
1592 | dm_table_presuspend_targets(map); | |
1593 | ||
bfa152fa JN |
1594 | /* bdget() can stall if the pending I/Os are not flushed */ |
1595 | if (!noflush) { | |
1596 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1597 | if (!md->suspended_bdev) { | |
1598 | DMWARN("bdget failed in dm_suspend"); | |
1599 | r = -ENOMEM; | |
f431d966 | 1600 | goto out; |
bfa152fa | 1601 | } |
e39e2e95 | 1602 | |
6d6f10df MB |
1603 | /* |
1604 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1605 | * because lock_fs() needs to flush I/Os. | |
1606 | */ | |
1607 | if (do_lockfs) { | |
1608 | r = lock_fs(md); | |
1609 | if (r) | |
1610 | goto out; | |
1611 | } | |
aa8d7c2f | 1612 | } |
1da177e4 LT |
1613 | |
1614 | /* | |
3b00b203 MP |
1615 | * Here we must make sure that no processes are submitting requests |
1616 | * to target drivers i.e. no one may be executing | |
1617 | * __split_and_process_bio. This is called from dm_request and | |
1618 | * dm_wq_work. | |
1619 | * | |
1620 | * To get all processes out of __split_and_process_bio in dm_request, | |
1621 | * we take the write lock. To prevent any process from reentering | |
1622 | * __split_and_process_bio from dm_request, we set | |
1623 | * DMF_QUEUE_IO_TO_THREAD. | |
1624 | * | |
1625 | * To quiesce the thread (dm_wq_work), we set DMF_BLOCK_IO_FOR_SUSPEND | |
1626 | * and call flush_workqueue(md->wq). flush_workqueue will wait until | |
1627 | * dm_wq_work exits and DMF_BLOCK_IO_FOR_SUSPEND will prevent any | |
1628 | * further calls to __split_and_process_bio from dm_wq_work. | |
1da177e4 | 1629 | */ |
2ca3310e | 1630 | down_write(&md->io_lock); |
1eb787ec AK |
1631 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1632 | set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); | |
2ca3310e | 1633 | up_write(&md->io_lock); |
1da177e4 | 1634 | |
3b00b203 MP |
1635 | flush_workqueue(md->wq); |
1636 | ||
1da177e4 | 1637 | /* |
3b00b203 MP |
1638 | * At this point no more requests are entering target request routines. |
1639 | * We call dm_wait_for_completion to wait for all existing requests | |
1640 | * to finish. | |
1da177e4 | 1641 | */ |
401600df | 1642 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 1643 | |
2ca3310e | 1644 | down_write(&md->io_lock); |
6d6f10df | 1645 | if (noflush) |
022c2611 | 1646 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 1647 | up_write(&md->io_lock); |
2e93ccc1 | 1648 | |
1da177e4 | 1649 | /* were we interrupted ? */ |
46125c1c | 1650 | if (r < 0) { |
9a1fb464 | 1651 | dm_queue_flush(md); |
73d410c0 | 1652 | |
2ca3310e | 1653 | unlock_fs(md); |
2e93ccc1 | 1654 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1655 | } |
1da177e4 | 1656 | |
3b00b203 MP |
1657 | /* |
1658 | * If dm_wait_for_completion returned 0, the device is completely | |
1659 | * quiescent now. There is no request-processing activity. All new | |
1660 | * requests are being added to md->deferred list. | |
1661 | */ | |
1662 | ||
cf222b37 | 1663 | dm_table_postsuspend_targets(map); |
1da177e4 | 1664 | |
2ca3310e | 1665 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1666 | |
2ca3310e | 1667 | out: |
e39e2e95 AK |
1668 | if (r && md->suspended_bdev) { |
1669 | bdput(md->suspended_bdev); | |
1670 | md->suspended_bdev = NULL; | |
1671 | } | |
1672 | ||
2ca3310e | 1673 | dm_table_put(map); |
d287483d AK |
1674 | |
1675 | out_unlock: | |
e61290a4 | 1676 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1677 | return r; |
1da177e4 LT |
1678 | } |
1679 | ||
1680 | int dm_resume(struct mapped_device *md) | |
1681 | { | |
cf222b37 | 1682 | int r = -EINVAL; |
cf222b37 | 1683 | struct dm_table *map = NULL; |
1da177e4 | 1684 | |
e61290a4 | 1685 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1686 | if (!dm_suspended(md)) |
cf222b37 | 1687 | goto out; |
cf222b37 AK |
1688 | |
1689 | map = dm_get_table(md); | |
2ca3310e | 1690 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1691 | goto out; |
1da177e4 | 1692 | |
8757b776 MB |
1693 | r = dm_table_resume_targets(map); |
1694 | if (r) | |
1695 | goto out; | |
2ca3310e | 1696 | |
9a1fb464 | 1697 | dm_queue_flush(md); |
2ca3310e AK |
1698 | |
1699 | unlock_fs(md); | |
1700 | ||
bfa152fa JN |
1701 | if (md->suspended_bdev) { |
1702 | bdput(md->suspended_bdev); | |
1703 | md->suspended_bdev = NULL; | |
1704 | } | |
e39e2e95 | 1705 | |
2ca3310e AK |
1706 | clear_bit(DMF_SUSPENDED, &md->flags); |
1707 | ||
1da177e4 | 1708 | dm_table_unplug_all(map); |
1da177e4 | 1709 | |
69267a30 | 1710 | dm_kobject_uevent(md); |
8560ed6f | 1711 | |
cf222b37 | 1712 | r = 0; |
2ca3310e | 1713 | |
cf222b37 AK |
1714 | out: |
1715 | dm_table_put(map); | |
e61290a4 | 1716 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1717 | |
cf222b37 | 1718 | return r; |
1da177e4 LT |
1719 | } |
1720 | ||
1721 | /*----------------------------------------------------------------- | |
1722 | * Event notification. | |
1723 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1724 | void dm_kobject_uevent(struct mapped_device *md) |
1725 | { | |
ed9e1982 | 1726 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1727 | } |
1728 | ||
7a8c3d3b MA |
1729 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1730 | { | |
1731 | return atomic_add_return(1, &md->uevent_seq); | |
1732 | } | |
1733 | ||
1da177e4 LT |
1734 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1735 | { | |
1736 | return atomic_read(&md->event_nr); | |
1737 | } | |
1738 | ||
1739 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1740 | { | |
1741 | return wait_event_interruptible(md->eventq, | |
1742 | (event_nr != atomic_read(&md->event_nr))); | |
1743 | } | |
1744 | ||
7a8c3d3b MA |
1745 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1746 | { | |
1747 | unsigned long flags; | |
1748 | ||
1749 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1750 | list_add(elist, &md->uevent_list); | |
1751 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1752 | } | |
1753 | ||
1da177e4 LT |
1754 | /* |
1755 | * The gendisk is only valid as long as you have a reference | |
1756 | * count on 'md'. | |
1757 | */ | |
1758 | struct gendisk *dm_disk(struct mapped_device *md) | |
1759 | { | |
1760 | return md->disk; | |
1761 | } | |
1762 | ||
784aae73 MB |
1763 | struct kobject *dm_kobject(struct mapped_device *md) |
1764 | { | |
1765 | return &md->kobj; | |
1766 | } | |
1767 | ||
1768 | /* | |
1769 | * struct mapped_device should not be exported outside of dm.c | |
1770 | * so use this check to verify that kobj is part of md structure | |
1771 | */ | |
1772 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1773 | { | |
1774 | struct mapped_device *md; | |
1775 | ||
1776 | md = container_of(kobj, struct mapped_device, kobj); | |
1777 | if (&md->kobj != kobj) | |
1778 | return NULL; | |
1779 | ||
4d89b7b4 MB |
1780 | if (test_bit(DMF_FREEING, &md->flags) || |
1781 | test_bit(DMF_DELETING, &md->flags)) | |
1782 | return NULL; | |
1783 | ||
784aae73 MB |
1784 | dm_get(md); |
1785 | return md; | |
1786 | } | |
1787 | ||
1da177e4 LT |
1788 | int dm_suspended(struct mapped_device *md) |
1789 | { | |
1790 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1791 | } | |
1792 | ||
2e93ccc1 KU |
1793 | int dm_noflush_suspending(struct dm_target *ti) |
1794 | { | |
1795 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1796 | int r = __noflush_suspending(md); | |
1797 | ||
1798 | dm_put(md); | |
1799 | ||
1800 | return r; | |
1801 | } | |
1802 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1803 | ||
1da177e4 LT |
1804 | static struct block_device_operations dm_blk_dops = { |
1805 | .open = dm_blk_open, | |
1806 | .release = dm_blk_close, | |
aa129a22 | 1807 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1808 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1809 | .owner = THIS_MODULE |
1810 | }; | |
1811 | ||
1812 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1813 | ||
1814 | /* | |
1815 | * module hooks | |
1816 | */ | |
1817 | module_init(dm_init); | |
1818 | module_exit(dm_exit); | |
1819 | ||
1820 | module_param(major, uint, 0); | |
1821 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1822 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1823 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1824 | MODULE_LICENSE("GPL"); |