]>
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" | |
9 | #include "dm-bio-list.h" | |
51e5b2bd | 10 | #include "dm-uevent.h" |
1da177e4 LT |
11 | |
12 | #include <linux/init.h> | |
13 | #include <linux/module.h> | |
48c9c27b | 14 | #include <linux/mutex.h> |
1da177e4 LT |
15 | #include <linux/moduleparam.h> |
16 | #include <linux/blkpg.h> | |
17 | #include <linux/bio.h> | |
18 | #include <linux/buffer_head.h> | |
19 | #include <linux/mempool.h> | |
20 | #include <linux/slab.h> | |
21 | #include <linux/idr.h> | |
3ac51e74 | 22 | #include <linux/hdreg.h> |
2056a782 | 23 | #include <linux/blktrace_api.h> |
5f3ea37c | 24 | #include <trace/block.h> |
1da177e4 | 25 | |
72d94861 AK |
26 | #define DM_MSG_PREFIX "core" |
27 | ||
1da177e4 LT |
28 | static const char *_name = DM_NAME; |
29 | ||
30 | static unsigned int major = 0; | |
31 | static unsigned int _major = 0; | |
32 | ||
f32c10b0 | 33 | static DEFINE_SPINLOCK(_minor_lock); |
1da177e4 | 34 | /* |
8fbf26ad | 35 | * For bio-based dm. |
1da177e4 LT |
36 | * One of these is allocated per bio. |
37 | */ | |
38 | struct dm_io { | |
39 | struct mapped_device *md; | |
40 | int error; | |
1da177e4 | 41 | atomic_t io_count; |
6ae2fa67 | 42 | struct bio *bio; |
3eaf840e | 43 | unsigned long start_time; |
1da177e4 LT |
44 | }; |
45 | ||
46 | /* | |
8fbf26ad | 47 | * For bio-based dm. |
1da177e4 LT |
48 | * One of these is allocated per target within a bio. Hopefully |
49 | * this will be simplified out one day. | |
50 | */ | |
028867ac | 51 | struct dm_target_io { |
1da177e4 LT |
52 | struct dm_io *io; |
53 | struct dm_target *ti; | |
54 | union map_info info; | |
55 | }; | |
56 | ||
0bfc2455 IM |
57 | DEFINE_TRACE(block_bio_complete); |
58 | ||
8fbf26ad KU |
59 | /* |
60 | * For request-based dm. | |
61 | * One of these is allocated per request. | |
62 | */ | |
63 | struct dm_rq_target_io { | |
64 | struct mapped_device *md; | |
65 | struct dm_target *ti; | |
66 | struct request *orig, clone; | |
67 | int error; | |
68 | union map_info info; | |
69 | }; | |
70 | ||
71 | /* | |
72 | * For request-based dm. | |
73 | * One of these is allocated per bio. | |
74 | */ | |
75 | struct dm_rq_clone_bio_info { | |
76 | struct bio *orig; | |
77 | struct request *rq; | |
78 | }; | |
79 | ||
1da177e4 LT |
80 | union map_info *dm_get_mapinfo(struct bio *bio) |
81 | { | |
17b2f66f | 82 | if (bio && bio->bi_private) |
028867ac | 83 | return &((struct dm_target_io *)bio->bi_private)->info; |
17b2f66f | 84 | return NULL; |
1da177e4 LT |
85 | } |
86 | ||
ba61fdd1 JM |
87 | #define MINOR_ALLOCED ((void *)-1) |
88 | ||
1da177e4 LT |
89 | /* |
90 | * Bits for the md->flags field. | |
91 | */ | |
92 | #define DMF_BLOCK_IO 0 | |
93 | #define DMF_SUSPENDED 1 | |
aa8d7c2f | 94 | #define DMF_FROZEN 2 |
fba9f90e | 95 | #define DMF_FREEING 3 |
5c6bd75d | 96 | #define DMF_DELETING 4 |
2e93ccc1 | 97 | #define DMF_NOFLUSH_SUSPENDING 5 |
1da177e4 | 98 | |
304f3f6a MB |
99 | /* |
100 | * Work processed by per-device workqueue. | |
101 | */ | |
102 | struct dm_wq_req { | |
304f3f6a MB |
103 | struct work_struct work; |
104 | struct mapped_device *md; | |
105 | void *context; | |
106 | }; | |
107 | ||
1da177e4 | 108 | struct mapped_device { |
2ca3310e | 109 | struct rw_semaphore io_lock; |
e61290a4 | 110 | struct mutex suspend_lock; |
2e93ccc1 | 111 | spinlock_t pushback_lock; |
1da177e4 LT |
112 | rwlock_t map_lock; |
113 | atomic_t holders; | |
5c6bd75d | 114 | atomic_t open_count; |
1da177e4 LT |
115 | |
116 | unsigned long flags; | |
117 | ||
165125e1 | 118 | struct request_queue *queue; |
1da177e4 | 119 | struct gendisk *disk; |
7e51f257 | 120 | char name[16]; |
1da177e4 LT |
121 | |
122 | void *interface_ptr; | |
123 | ||
124 | /* | |
125 | * A list of ios that arrived while we were suspended. | |
126 | */ | |
127 | atomic_t pending; | |
128 | wait_queue_head_t wait; | |
74859364 | 129 | struct bio_list deferred; |
2e93ccc1 | 130 | struct bio_list pushback; |
1da177e4 | 131 | |
304f3f6a MB |
132 | /* |
133 | * Processing queue (flush/barriers) | |
134 | */ | |
135 | struct workqueue_struct *wq; | |
136 | ||
1da177e4 LT |
137 | /* |
138 | * The current mapping. | |
139 | */ | |
140 | struct dm_table *map; | |
141 | ||
142 | /* | |
143 | * io objects are allocated from here. | |
144 | */ | |
145 | mempool_t *io_pool; | |
146 | mempool_t *tio_pool; | |
147 | ||
9faf400f SB |
148 | struct bio_set *bs; |
149 | ||
1da177e4 LT |
150 | /* |
151 | * Event handling. | |
152 | */ | |
153 | atomic_t event_nr; | |
154 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
155 | atomic_t uevent_seq; |
156 | struct list_head uevent_list; | |
157 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
158 | |
159 | /* | |
160 | * freeze/thaw support require holding onto a super block | |
161 | */ | |
162 | struct super_block *frozen_sb; | |
e39e2e95 | 163 | struct block_device *suspended_bdev; |
3ac51e74 DW |
164 | |
165 | /* forced geometry settings */ | |
166 | struct hd_geometry geometry; | |
784aae73 MB |
167 | |
168 | /* sysfs handle */ | |
169 | struct kobject kobj; | |
1da177e4 LT |
170 | }; |
171 | ||
172 | #define MIN_IOS 256 | |
e18b890b CL |
173 | static struct kmem_cache *_io_cache; |
174 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
175 | static struct kmem_cache *_rq_tio_cache; |
176 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 177 | |
1da177e4 LT |
178 | static int __init local_init(void) |
179 | { | |
51157b4a | 180 | int r = -ENOMEM; |
1da177e4 | 181 | |
1da177e4 | 182 | /* allocate a slab for the dm_ios */ |
028867ac | 183 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 184 | if (!_io_cache) |
51157b4a | 185 | return r; |
1da177e4 LT |
186 | |
187 | /* allocate a slab for the target ios */ | |
028867ac | 188 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
189 | if (!_tio_cache) |
190 | goto out_free_io_cache; | |
1da177e4 | 191 | |
8fbf26ad KU |
192 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
193 | if (!_rq_tio_cache) | |
194 | goto out_free_tio_cache; | |
195 | ||
196 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
197 | if (!_rq_bio_info_cache) | |
198 | goto out_free_rq_tio_cache; | |
199 | ||
51e5b2bd | 200 | r = dm_uevent_init(); |
51157b4a | 201 | if (r) |
8fbf26ad | 202 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 203 | |
1da177e4 LT |
204 | _major = major; |
205 | r = register_blkdev(_major, _name); | |
51157b4a KU |
206 | if (r < 0) |
207 | goto out_uevent_exit; | |
1da177e4 LT |
208 | |
209 | if (!_major) | |
210 | _major = r; | |
211 | ||
212 | return 0; | |
51157b4a KU |
213 | |
214 | out_uevent_exit: | |
215 | dm_uevent_exit(); | |
8fbf26ad KU |
216 | out_free_rq_bio_info_cache: |
217 | kmem_cache_destroy(_rq_bio_info_cache); | |
218 | out_free_rq_tio_cache: | |
219 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
220 | out_free_tio_cache: |
221 | kmem_cache_destroy(_tio_cache); | |
222 | out_free_io_cache: | |
223 | kmem_cache_destroy(_io_cache); | |
224 | ||
225 | return r; | |
1da177e4 LT |
226 | } |
227 | ||
228 | static void local_exit(void) | |
229 | { | |
8fbf26ad KU |
230 | kmem_cache_destroy(_rq_bio_info_cache); |
231 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
232 | kmem_cache_destroy(_tio_cache); |
233 | kmem_cache_destroy(_io_cache); | |
00d59405 | 234 | unregister_blkdev(_major, _name); |
51e5b2bd | 235 | dm_uevent_exit(); |
1da177e4 LT |
236 | |
237 | _major = 0; | |
238 | ||
239 | DMINFO("cleaned up"); | |
240 | } | |
241 | ||
b9249e55 | 242 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
243 | local_init, |
244 | dm_target_init, | |
245 | dm_linear_init, | |
246 | dm_stripe_init, | |
945fa4d2 | 247 | dm_kcopyd_init, |
1da177e4 LT |
248 | dm_interface_init, |
249 | }; | |
250 | ||
b9249e55 | 251 | static void (*_exits[])(void) = { |
1da177e4 LT |
252 | local_exit, |
253 | dm_target_exit, | |
254 | dm_linear_exit, | |
255 | dm_stripe_exit, | |
945fa4d2 | 256 | dm_kcopyd_exit, |
1da177e4 LT |
257 | dm_interface_exit, |
258 | }; | |
259 | ||
260 | static int __init dm_init(void) | |
261 | { | |
262 | const int count = ARRAY_SIZE(_inits); | |
263 | ||
264 | int r, i; | |
265 | ||
266 | for (i = 0; i < count; i++) { | |
267 | r = _inits[i](); | |
268 | if (r) | |
269 | goto bad; | |
270 | } | |
271 | ||
272 | return 0; | |
273 | ||
274 | bad: | |
275 | while (i--) | |
276 | _exits[i](); | |
277 | ||
278 | return r; | |
279 | } | |
280 | ||
281 | static void __exit dm_exit(void) | |
282 | { | |
283 | int i = ARRAY_SIZE(_exits); | |
284 | ||
285 | while (i--) | |
286 | _exits[i](); | |
287 | } | |
288 | ||
289 | /* | |
290 | * Block device functions | |
291 | */ | |
fe5f9f2c | 292 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
293 | { |
294 | struct mapped_device *md; | |
295 | ||
fba9f90e JM |
296 | spin_lock(&_minor_lock); |
297 | ||
fe5f9f2c | 298 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
299 | if (!md) |
300 | goto out; | |
301 | ||
5c6bd75d AK |
302 | if (test_bit(DMF_FREEING, &md->flags) || |
303 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
304 | md = NULL; |
305 | goto out; | |
306 | } | |
307 | ||
1da177e4 | 308 | dm_get(md); |
5c6bd75d | 309 | atomic_inc(&md->open_count); |
fba9f90e JM |
310 | |
311 | out: | |
312 | spin_unlock(&_minor_lock); | |
313 | ||
314 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
315 | } |
316 | ||
fe5f9f2c | 317 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 318 | { |
fe5f9f2c | 319 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 320 | atomic_dec(&md->open_count); |
1da177e4 LT |
321 | dm_put(md); |
322 | return 0; | |
323 | } | |
324 | ||
5c6bd75d AK |
325 | int dm_open_count(struct mapped_device *md) |
326 | { | |
327 | return atomic_read(&md->open_count); | |
328 | } | |
329 | ||
330 | /* | |
331 | * Guarantees nothing is using the device before it's deleted. | |
332 | */ | |
333 | int dm_lock_for_deletion(struct mapped_device *md) | |
334 | { | |
335 | int r = 0; | |
336 | ||
337 | spin_lock(&_minor_lock); | |
338 | ||
339 | if (dm_open_count(md)) | |
340 | r = -EBUSY; | |
341 | else | |
342 | set_bit(DMF_DELETING, &md->flags); | |
343 | ||
344 | spin_unlock(&_minor_lock); | |
345 | ||
346 | return r; | |
347 | } | |
348 | ||
3ac51e74 DW |
349 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
350 | { | |
351 | struct mapped_device *md = bdev->bd_disk->private_data; | |
352 | ||
353 | return dm_get_geometry(md, geo); | |
354 | } | |
355 | ||
fe5f9f2c | 356 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
357 | unsigned int cmd, unsigned long arg) |
358 | { | |
fe5f9f2c AV |
359 | struct mapped_device *md = bdev->bd_disk->private_data; |
360 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
361 | struct dm_target *tgt; |
362 | int r = -ENOTTY; | |
363 | ||
aa129a22 MB |
364 | if (!map || !dm_table_get_size(map)) |
365 | goto out; | |
366 | ||
367 | /* We only support devices that have a single target */ | |
368 | if (dm_table_get_num_targets(map) != 1) | |
369 | goto out; | |
370 | ||
371 | tgt = dm_table_get_target(map, 0); | |
372 | ||
373 | if (dm_suspended(md)) { | |
374 | r = -EAGAIN; | |
375 | goto out; | |
376 | } | |
377 | ||
378 | if (tgt->type->ioctl) | |
647b3d00 | 379 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
380 | |
381 | out: | |
382 | dm_table_put(map); | |
383 | ||
aa129a22 MB |
384 | return r; |
385 | } | |
386 | ||
028867ac | 387 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
388 | { |
389 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
390 | } | |
391 | ||
028867ac | 392 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
393 | { |
394 | mempool_free(io, md->io_pool); | |
395 | } | |
396 | ||
028867ac | 397 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
398 | { |
399 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
400 | } | |
401 | ||
028867ac | 402 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
403 | { |
404 | mempool_free(tio, md->tio_pool); | |
405 | } | |
406 | ||
3eaf840e JNN |
407 | static void start_io_acct(struct dm_io *io) |
408 | { | |
409 | struct mapped_device *md = io->md; | |
c9959059 | 410 | int cpu; |
3eaf840e JNN |
411 | |
412 | io->start_time = jiffies; | |
413 | ||
074a7aca TH |
414 | cpu = part_stat_lock(); |
415 | part_round_stats(cpu, &dm_disk(md)->part0); | |
416 | part_stat_unlock(); | |
417 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
418 | } |
419 | ||
d221d2e7 | 420 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
421 | { |
422 | struct mapped_device *md = io->md; | |
423 | struct bio *bio = io->bio; | |
424 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 425 | int pending, cpu; |
3eaf840e JNN |
426 | int rw = bio_data_dir(bio); |
427 | ||
074a7aca TH |
428 | cpu = part_stat_lock(); |
429 | part_round_stats(cpu, &dm_disk(md)->part0); | |
430 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
431 | part_stat_unlock(); | |
3eaf840e | 432 | |
074a7aca TH |
433 | dm_disk(md)->part0.in_flight = pending = |
434 | atomic_dec_return(&md->pending); | |
3eaf840e | 435 | |
d221d2e7 MP |
436 | /* nudge anyone waiting on suspend queue */ |
437 | if (!pending) | |
438 | wake_up(&md->wait); | |
3eaf840e JNN |
439 | } |
440 | ||
1da177e4 LT |
441 | /* |
442 | * Add the bio to the list of deferred io. | |
443 | */ | |
444 | static int queue_io(struct mapped_device *md, struct bio *bio) | |
445 | { | |
2ca3310e | 446 | down_write(&md->io_lock); |
1da177e4 LT |
447 | |
448 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 449 | up_write(&md->io_lock); |
1da177e4 LT |
450 | return 1; |
451 | } | |
452 | ||
453 | bio_list_add(&md->deferred, bio); | |
454 | ||
2ca3310e | 455 | up_write(&md->io_lock); |
1da177e4 LT |
456 | return 0; /* deferred successfully */ |
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. | |
537 | * This must be handled before the sleeper on | |
538 | * suspend queue merges the pushback list. | |
539 | */ | |
b35f8caa MB |
540 | spin_lock_irqsave(&md->pushback_lock, flags); |
541 | if (__noflush_suspending(md)) | |
542 | bio_list_add(&md->pushback, io->bio); | |
2e93ccc1 KU |
543 | else |
544 | /* noflush suspend was interrupted. */ | |
545 | io->error = -EIO; | |
b35f8caa | 546 | spin_unlock_irqrestore(&md->pushback_lock, flags); |
2e93ccc1 KU |
547 | } |
548 | ||
d221d2e7 | 549 | end_io_acct(io); |
1da177e4 | 550 | |
b35f8caa MB |
551 | io_error = io->error; |
552 | bio = io->bio; | |
2e93ccc1 | 553 | |
b35f8caa MB |
554 | free_io(md, io); |
555 | ||
556 | if (io_error != DM_ENDIO_REQUEUE) { | |
557 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 558 | |
b35f8caa MB |
559 | bio_endio(bio, io_error); |
560 | } | |
1da177e4 LT |
561 | } |
562 | } | |
563 | ||
6712ecf8 | 564 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
565 | { |
566 | int r = 0; | |
028867ac | 567 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 568 | struct dm_io *io = tio->io; |
9faf400f | 569 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
570 | dm_endio_fn endio = tio->ti->type->end_io; |
571 | ||
1da177e4 LT |
572 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
573 | error = -EIO; | |
574 | ||
575 | if (endio) { | |
576 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
577 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
578 | /* | |
579 | * error and requeue request are handled | |
580 | * in dec_pending(). | |
581 | */ | |
1da177e4 | 582 | error = r; |
45cbcd79 KU |
583 | else if (r == DM_ENDIO_INCOMPLETE) |
584 | /* The target will handle the io */ | |
6712ecf8 | 585 | return; |
45cbcd79 KU |
586 | else if (r) { |
587 | DMWARN("unimplemented target endio return value: %d", r); | |
588 | BUG(); | |
589 | } | |
1da177e4 LT |
590 | } |
591 | ||
9faf400f SB |
592 | /* |
593 | * Store md for cleanup instead of tio which is about to get freed. | |
594 | */ | |
595 | bio->bi_private = md->bs; | |
596 | ||
9faf400f | 597 | free_tio(md, tio); |
b35f8caa MB |
598 | bio_put(bio); |
599 | dec_pending(io, error); | |
1da177e4 LT |
600 | } |
601 | ||
602 | static sector_t max_io_len(struct mapped_device *md, | |
603 | sector_t sector, struct dm_target *ti) | |
604 | { | |
605 | sector_t offset = sector - ti->begin; | |
606 | sector_t len = ti->len - offset; | |
607 | ||
608 | /* | |
609 | * Does the target need to split even further ? | |
610 | */ | |
611 | if (ti->split_io) { | |
612 | sector_t boundary; | |
613 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
614 | - offset; | |
615 | if (len > boundary) | |
616 | len = boundary; | |
617 | } | |
618 | ||
619 | return len; | |
620 | } | |
621 | ||
622 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 623 | struct dm_target_io *tio) |
1da177e4 LT |
624 | { |
625 | int r; | |
2056a782 | 626 | sector_t sector; |
9faf400f | 627 | struct mapped_device *md; |
1da177e4 LT |
628 | |
629 | /* | |
630 | * Sanity checks. | |
631 | */ | |
632 | BUG_ON(!clone->bi_size); | |
633 | ||
634 | clone->bi_end_io = clone_endio; | |
635 | clone->bi_private = tio; | |
636 | ||
637 | /* | |
638 | * Map the clone. If r == 0 we don't need to do | |
639 | * anything, the target has assumed ownership of | |
640 | * this io. | |
641 | */ | |
642 | atomic_inc(&tio->io->io_count); | |
2056a782 | 643 | sector = clone->bi_sector; |
1da177e4 | 644 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 645 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 646 | /* the bio has been remapped so dispatch it */ |
2056a782 | 647 | |
5f3ea37c | 648 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
649 | tio->io->bio->bi_bdev->bd_dev, |
650 | clone->bi_sector, sector); | |
2056a782 | 651 | |
1da177e4 | 652 | generic_make_request(clone); |
2e93ccc1 KU |
653 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
654 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
655 | md = tio->io->md; |
656 | dec_pending(tio->io, r); | |
657 | /* | |
658 | * Store bio_set for cleanup. | |
659 | */ | |
660 | clone->bi_private = md->bs; | |
1da177e4 | 661 | bio_put(clone); |
9faf400f | 662 | free_tio(md, tio); |
45cbcd79 KU |
663 | } else if (r) { |
664 | DMWARN("unimplemented target map return value: %d", r); | |
665 | BUG(); | |
1da177e4 LT |
666 | } |
667 | } | |
668 | ||
669 | struct clone_info { | |
670 | struct mapped_device *md; | |
671 | struct dm_table *map; | |
672 | struct bio *bio; | |
673 | struct dm_io *io; | |
674 | sector_t sector; | |
675 | sector_t sector_count; | |
676 | unsigned short idx; | |
677 | }; | |
678 | ||
3676347a PO |
679 | static void dm_bio_destructor(struct bio *bio) |
680 | { | |
9faf400f SB |
681 | struct bio_set *bs = bio->bi_private; |
682 | ||
683 | bio_free(bio, bs); | |
3676347a PO |
684 | } |
685 | ||
1da177e4 LT |
686 | /* |
687 | * Creates a little bio that is just does part of a bvec. | |
688 | */ | |
689 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
690 | unsigned short idx, unsigned int offset, | |
9faf400f | 691 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
692 | { |
693 | struct bio *clone; | |
694 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
695 | ||
9faf400f | 696 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 697 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
698 | *clone->bi_io_vec = *bv; |
699 | ||
700 | clone->bi_sector = sector; | |
701 | clone->bi_bdev = bio->bi_bdev; | |
702 | clone->bi_rw = bio->bi_rw; | |
703 | clone->bi_vcnt = 1; | |
704 | clone->bi_size = to_bytes(len); | |
705 | clone->bi_io_vec->bv_offset = offset; | |
706 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 707 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 LT |
708 | |
709 | return clone; | |
710 | } | |
711 | ||
712 | /* | |
713 | * Creates a bio that consists of range of complete bvecs. | |
714 | */ | |
715 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
716 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 717 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
718 | { |
719 | struct bio *clone; | |
720 | ||
9faf400f SB |
721 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
722 | __bio_clone(clone, bio); | |
723 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
724 | clone->bi_sector = sector; |
725 | clone->bi_idx = idx; | |
726 | clone->bi_vcnt = idx + bv_count; | |
727 | clone->bi_size = to_bytes(len); | |
728 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
729 | ||
730 | return clone; | |
731 | } | |
732 | ||
512875bd | 733 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
734 | { |
735 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
736 | struct dm_target *ti; |
737 | sector_t len = 0, max; | |
028867ac | 738 | struct dm_target_io *tio; |
1da177e4 | 739 | |
512875bd JN |
740 | ti = dm_table_find_target(ci->map, ci->sector); |
741 | if (!dm_target_is_valid(ti)) | |
742 | return -EIO; | |
743 | ||
744 | max = max_io_len(ci->md, ci->sector, ti); | |
745 | ||
1da177e4 LT |
746 | /* |
747 | * Allocate a target io object. | |
748 | */ | |
749 | tio = alloc_tio(ci->md); | |
750 | tio->io = ci->io; | |
751 | tio->ti = ti; | |
752 | memset(&tio->info, 0, sizeof(tio->info)); | |
753 | ||
754 | if (ci->sector_count <= max) { | |
755 | /* | |
756 | * Optimise for the simple case where we can do all of | |
757 | * the remaining io with a single clone. | |
758 | */ | |
759 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
760 | bio->bi_vcnt - ci->idx, ci->sector_count, |
761 | ci->md->bs); | |
1da177e4 LT |
762 | __map_bio(ti, clone, tio); |
763 | ci->sector_count = 0; | |
764 | ||
765 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
766 | /* | |
767 | * There are some bvecs that don't span targets. | |
768 | * Do as many of these as possible. | |
769 | */ | |
770 | int i; | |
771 | sector_t remaining = max; | |
772 | sector_t bv_len; | |
773 | ||
774 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
775 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
776 | ||
777 | if (bv_len > remaining) | |
778 | break; | |
779 | ||
780 | remaining -= bv_len; | |
781 | len += bv_len; | |
782 | } | |
783 | ||
9faf400f SB |
784 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
785 | ci->md->bs); | |
1da177e4 LT |
786 | __map_bio(ti, clone, tio); |
787 | ||
788 | ci->sector += len; | |
789 | ci->sector_count -= len; | |
790 | ci->idx = i; | |
791 | ||
792 | } else { | |
793 | /* | |
d2044a94 | 794 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
795 | */ |
796 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
797 | sector_t remaining = to_sector(bv->bv_len); |
798 | unsigned int offset = 0; | |
1da177e4 | 799 | |
d2044a94 AK |
800 | do { |
801 | if (offset) { | |
802 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
803 | if (!dm_target_is_valid(ti)) |
804 | return -EIO; | |
805 | ||
d2044a94 | 806 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 807 | |
d2044a94 AK |
808 | tio = alloc_tio(ci->md); |
809 | tio->io = ci->io; | |
810 | tio->ti = ti; | |
811 | memset(&tio->info, 0, sizeof(tio->info)); | |
812 | } | |
813 | ||
814 | len = min(remaining, max); | |
815 | ||
816 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
817 | bv->bv_offset + offset, len, |
818 | ci->md->bs); | |
d2044a94 AK |
819 | |
820 | __map_bio(ti, clone, tio); | |
821 | ||
822 | ci->sector += len; | |
823 | ci->sector_count -= len; | |
824 | offset += to_bytes(len); | |
825 | } while (remaining -= len); | |
1da177e4 | 826 | |
1da177e4 LT |
827 | ci->idx++; |
828 | } | |
512875bd JN |
829 | |
830 | return 0; | |
1da177e4 LT |
831 | } |
832 | ||
833 | /* | |
834 | * Split the bio into several clones. | |
835 | */ | |
9e4e5f87 | 836 | static int __split_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
837 | { |
838 | struct clone_info ci; | |
512875bd | 839 | int error = 0; |
1da177e4 LT |
840 | |
841 | ci.map = dm_get_table(md); | |
9e4e5f87 MB |
842 | if (unlikely(!ci.map)) |
843 | return -EIO; | |
ab4c1424 AK |
844 | if (unlikely(bio_barrier(bio) && !dm_table_barrier_ok(ci.map))) { |
845 | dm_table_put(ci.map); | |
846 | bio_endio(bio, -EOPNOTSUPP); | |
847 | return 0; | |
848 | } | |
1da177e4 LT |
849 | ci.md = md; |
850 | ci.bio = bio; | |
851 | ci.io = alloc_io(md); | |
852 | ci.io->error = 0; | |
853 | atomic_set(&ci.io->io_count, 1); | |
854 | ci.io->bio = bio; | |
855 | ci.io->md = md; | |
856 | ci.sector = bio->bi_sector; | |
857 | ci.sector_count = bio_sectors(bio); | |
858 | ci.idx = bio->bi_idx; | |
859 | ||
3eaf840e | 860 | start_io_acct(ci.io); |
512875bd JN |
861 | while (ci.sector_count && !error) |
862 | error = __clone_and_map(&ci); | |
1da177e4 LT |
863 | |
864 | /* drop the extra reference count */ | |
512875bd | 865 | dec_pending(ci.io, error); |
1da177e4 | 866 | dm_table_put(ci.map); |
9e4e5f87 MB |
867 | |
868 | return 0; | |
1da177e4 LT |
869 | } |
870 | /*----------------------------------------------------------------- | |
871 | * CRUD END | |
872 | *---------------------------------------------------------------*/ | |
873 | ||
f6fccb12 MB |
874 | static int dm_merge_bvec(struct request_queue *q, |
875 | struct bvec_merge_data *bvm, | |
876 | struct bio_vec *biovec) | |
877 | { | |
878 | struct mapped_device *md = q->queuedata; | |
879 | struct dm_table *map = dm_get_table(md); | |
880 | struct dm_target *ti; | |
881 | sector_t max_sectors; | |
5037108a | 882 | int max_size = 0; |
f6fccb12 MB |
883 | |
884 | if (unlikely(!map)) | |
5037108a | 885 | goto out; |
f6fccb12 MB |
886 | |
887 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
888 | if (!dm_target_is_valid(ti)) |
889 | goto out_table; | |
f6fccb12 MB |
890 | |
891 | /* | |
892 | * Find maximum amount of I/O that won't need splitting | |
893 | */ | |
894 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
895 | (sector_t) BIO_MAX_SECTORS); | |
896 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
897 | if (max_size < 0) | |
898 | max_size = 0; | |
899 | ||
900 | /* | |
901 | * merge_bvec_fn() returns number of bytes | |
902 | * it can accept at this offset | |
903 | * max is precomputed maximal io size | |
904 | */ | |
905 | if (max_size && ti->type->merge) | |
906 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
907 | ||
b01cd5ac | 908 | out_table: |
5037108a MP |
909 | dm_table_put(map); |
910 | ||
911 | out: | |
f6fccb12 MB |
912 | /* |
913 | * Always allow an entire first page | |
914 | */ | |
915 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
916 | max_size = biovec->bv_len; | |
917 | ||
f6fccb12 MB |
918 | return max_size; |
919 | } | |
920 | ||
1da177e4 LT |
921 | /* |
922 | * The request function that just remaps the bio built up by | |
923 | * dm_merge_bvec. | |
924 | */ | |
165125e1 | 925 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 926 | { |
9e4e5f87 | 927 | int r = -EIO; |
12f03a49 | 928 | int rw = bio_data_dir(bio); |
1da177e4 | 929 | struct mapped_device *md = q->queuedata; |
c9959059 | 930 | int cpu; |
1da177e4 | 931 | |
2ca3310e | 932 | down_read(&md->io_lock); |
1da177e4 | 933 | |
074a7aca TH |
934 | cpu = part_stat_lock(); |
935 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
936 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
937 | part_stat_unlock(); | |
12f03a49 | 938 | |
1da177e4 LT |
939 | /* |
940 | * If we're suspended we have to queue | |
941 | * this io for later. | |
942 | */ | |
943 | while (test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 944 | up_read(&md->io_lock); |
1da177e4 | 945 | |
9e4e5f87 MB |
946 | if (bio_rw(bio) != READA) |
947 | r = queue_io(md, bio); | |
1da177e4 | 948 | |
9e4e5f87 MB |
949 | if (r <= 0) |
950 | goto out_req; | |
1da177e4 LT |
951 | |
952 | /* | |
953 | * We're in a while loop, because someone could suspend | |
954 | * before we get to the following read lock. | |
955 | */ | |
2ca3310e | 956 | down_read(&md->io_lock); |
1da177e4 LT |
957 | } |
958 | ||
9e4e5f87 | 959 | r = __split_bio(md, bio); |
2ca3310e | 960 | up_read(&md->io_lock); |
9e4e5f87 MB |
961 | |
962 | out_req: | |
963 | if (r < 0) | |
964 | bio_io_error(bio); | |
965 | ||
1da177e4 LT |
966 | return 0; |
967 | } | |
968 | ||
165125e1 | 969 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
970 | { |
971 | struct mapped_device *md = q->queuedata; | |
972 | struct dm_table *map = dm_get_table(md); | |
973 | ||
974 | if (map) { | |
975 | dm_table_unplug_all(map); | |
976 | dm_table_put(map); | |
977 | } | |
978 | } | |
979 | ||
980 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
981 | { | |
8a57dfc6 CS |
982 | int r = bdi_bits; |
983 | struct mapped_device *md = congested_data; | |
984 | struct dm_table *map; | |
1da177e4 | 985 | |
8a57dfc6 CS |
986 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { |
987 | map = dm_get_table(md); | |
988 | if (map) { | |
989 | r = dm_table_any_congested(map, bdi_bits); | |
990 | dm_table_put(map); | |
991 | } | |
992 | } | |
993 | ||
1da177e4 LT |
994 | return r; |
995 | } | |
996 | ||
997 | /*----------------------------------------------------------------- | |
998 | * An IDR is used to keep track of allocated minor numbers. | |
999 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1000 | static DEFINE_IDR(_minor_idr); |
1001 | ||
2b06cfff | 1002 | static void free_minor(int minor) |
1da177e4 | 1003 | { |
f32c10b0 | 1004 | spin_lock(&_minor_lock); |
1da177e4 | 1005 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1006 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1007 | } |
1008 | ||
1009 | /* | |
1010 | * See if the device with a specific minor # is free. | |
1011 | */ | |
cf13ab8e | 1012 | static int specific_minor(int minor) |
1da177e4 LT |
1013 | { |
1014 | int r, m; | |
1015 | ||
1016 | if (minor >= (1 << MINORBITS)) | |
1017 | return -EINVAL; | |
1018 | ||
62f75c2f JM |
1019 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1020 | if (!r) | |
1021 | return -ENOMEM; | |
1022 | ||
f32c10b0 | 1023 | spin_lock(&_minor_lock); |
1da177e4 LT |
1024 | |
1025 | if (idr_find(&_minor_idr, minor)) { | |
1026 | r = -EBUSY; | |
1027 | goto out; | |
1028 | } | |
1029 | ||
ba61fdd1 | 1030 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1031 | if (r) |
1da177e4 | 1032 | goto out; |
1da177e4 LT |
1033 | |
1034 | if (m != minor) { | |
1035 | idr_remove(&_minor_idr, m); | |
1036 | r = -EBUSY; | |
1037 | goto out; | |
1038 | } | |
1039 | ||
1040 | out: | |
f32c10b0 | 1041 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1042 | return r; |
1043 | } | |
1044 | ||
cf13ab8e | 1045 | static int next_free_minor(int *minor) |
1da177e4 | 1046 | { |
2b06cfff | 1047 | int r, m; |
1da177e4 | 1048 | |
1da177e4 | 1049 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1050 | if (!r) |
1051 | return -ENOMEM; | |
1052 | ||
f32c10b0 | 1053 | spin_lock(&_minor_lock); |
1da177e4 | 1054 | |
ba61fdd1 | 1055 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1056 | if (r) |
1da177e4 | 1057 | goto out; |
1da177e4 LT |
1058 | |
1059 | if (m >= (1 << MINORBITS)) { | |
1060 | idr_remove(&_minor_idr, m); | |
1061 | r = -ENOSPC; | |
1062 | goto out; | |
1063 | } | |
1064 | ||
1065 | *minor = m; | |
1066 | ||
1067 | out: | |
f32c10b0 | 1068 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1069 | return r; |
1070 | } | |
1071 | ||
1072 | static struct block_device_operations dm_blk_dops; | |
1073 | ||
1074 | /* | |
1075 | * Allocate and initialise a blank device with a given minor. | |
1076 | */ | |
2b06cfff | 1077 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1078 | { |
1079 | int r; | |
cf13ab8e | 1080 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1081 | void *old_md; |
1da177e4 LT |
1082 | |
1083 | if (!md) { | |
1084 | DMWARN("unable to allocate device, out of memory."); | |
1085 | return NULL; | |
1086 | } | |
1087 | ||
10da4f79 | 1088 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1089 | goto bad_module_get; |
10da4f79 | 1090 | |
1da177e4 | 1091 | /* get a minor number for the dev */ |
2b06cfff | 1092 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1093 | r = next_free_minor(&minor); |
2b06cfff | 1094 | else |
cf13ab8e | 1095 | r = specific_minor(minor); |
1da177e4 | 1096 | if (r < 0) |
6ed7ade8 | 1097 | goto bad_minor; |
1da177e4 | 1098 | |
2ca3310e | 1099 | init_rwsem(&md->io_lock); |
e61290a4 | 1100 | mutex_init(&md->suspend_lock); |
2e93ccc1 | 1101 | spin_lock_init(&md->pushback_lock); |
1da177e4 LT |
1102 | rwlock_init(&md->map_lock); |
1103 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1104 | atomic_set(&md->open_count, 0); |
1da177e4 | 1105 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1106 | atomic_set(&md->uevent_seq, 0); |
1107 | INIT_LIST_HEAD(&md->uevent_list); | |
1108 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1109 | |
1110 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1111 | if (!md->queue) | |
6ed7ade8 | 1112 | goto bad_queue; |
1da177e4 LT |
1113 | |
1114 | md->queue->queuedata = md; | |
1115 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1116 | md->queue->backing_dev_info.congested_data = md; | |
1117 | blk_queue_make_request(md->queue, dm_request); | |
daef265f | 1118 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1119 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1120 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1121 | |
93d2341c | 1122 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1123 | if (!md->io_pool) |
6ed7ade8 | 1124 | goto bad_io_pool; |
1da177e4 | 1125 | |
93d2341c | 1126 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1127 | if (!md->tio_pool) |
6ed7ade8 | 1128 | goto bad_tio_pool; |
1da177e4 | 1129 | |
bb799ca0 | 1130 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1131 | if (!md->bs) |
1132 | goto bad_no_bioset; | |
1133 | ||
1da177e4 LT |
1134 | md->disk = alloc_disk(1); |
1135 | if (!md->disk) | |
6ed7ade8 | 1136 | goto bad_disk; |
1da177e4 | 1137 | |
f0b04115 JM |
1138 | atomic_set(&md->pending, 0); |
1139 | init_waitqueue_head(&md->wait); | |
1140 | init_waitqueue_head(&md->eventq); | |
1141 | ||
1da177e4 LT |
1142 | md->disk->major = _major; |
1143 | md->disk->first_minor = minor; | |
1144 | md->disk->fops = &dm_blk_dops; | |
1145 | md->disk->queue = md->queue; | |
1146 | md->disk->private_data = md; | |
1147 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1148 | add_disk(md->disk); | |
7e51f257 | 1149 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1150 | |
304f3f6a MB |
1151 | md->wq = create_singlethread_workqueue("kdmflush"); |
1152 | if (!md->wq) | |
1153 | goto bad_thread; | |
1154 | ||
ba61fdd1 | 1155 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1156 | spin_lock(&_minor_lock); |
ba61fdd1 | 1157 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1158 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1159 | |
1160 | BUG_ON(old_md != MINOR_ALLOCED); | |
1161 | ||
1da177e4 LT |
1162 | return md; |
1163 | ||
304f3f6a MB |
1164 | bad_thread: |
1165 | put_disk(md->disk); | |
6ed7ade8 | 1166 | bad_disk: |
9faf400f | 1167 | bioset_free(md->bs); |
6ed7ade8 | 1168 | bad_no_bioset: |
1da177e4 | 1169 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1170 | bad_tio_pool: |
1da177e4 | 1171 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1172 | bad_io_pool: |
1312f40e | 1173 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1174 | bad_queue: |
1da177e4 | 1175 | free_minor(minor); |
6ed7ade8 | 1176 | bad_minor: |
10da4f79 | 1177 | module_put(THIS_MODULE); |
6ed7ade8 | 1178 | bad_module_get: |
1da177e4 LT |
1179 | kfree(md); |
1180 | return NULL; | |
1181 | } | |
1182 | ||
ae9da83f JN |
1183 | static void unlock_fs(struct mapped_device *md); |
1184 | ||
1da177e4 LT |
1185 | static void free_dev(struct mapped_device *md) |
1186 | { | |
f331c029 | 1187 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1188 | |
d9dde59b | 1189 | if (md->suspended_bdev) { |
ae9da83f | 1190 | unlock_fs(md); |
d9dde59b JN |
1191 | bdput(md->suspended_bdev); |
1192 | } | |
304f3f6a | 1193 | destroy_workqueue(md->wq); |
1da177e4 LT |
1194 | mempool_destroy(md->tio_pool); |
1195 | mempool_destroy(md->io_pool); | |
9faf400f | 1196 | bioset_free(md->bs); |
1da177e4 | 1197 | del_gendisk(md->disk); |
63d94e48 | 1198 | free_minor(minor); |
fba9f90e JM |
1199 | |
1200 | spin_lock(&_minor_lock); | |
1201 | md->disk->private_data = NULL; | |
1202 | spin_unlock(&_minor_lock); | |
1203 | ||
1da177e4 | 1204 | put_disk(md->disk); |
1312f40e | 1205 | blk_cleanup_queue(md->queue); |
10da4f79 | 1206 | module_put(THIS_MODULE); |
1da177e4 LT |
1207 | kfree(md); |
1208 | } | |
1209 | ||
1210 | /* | |
1211 | * Bind a table to the device. | |
1212 | */ | |
1213 | static void event_callback(void *context) | |
1214 | { | |
7a8c3d3b MA |
1215 | unsigned long flags; |
1216 | LIST_HEAD(uevents); | |
1da177e4 LT |
1217 | struct mapped_device *md = (struct mapped_device *) context; |
1218 | ||
7a8c3d3b MA |
1219 | spin_lock_irqsave(&md->uevent_lock, flags); |
1220 | list_splice_init(&md->uevent_list, &uevents); | |
1221 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1222 | ||
ed9e1982 | 1223 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1224 | |
1da177e4 LT |
1225 | atomic_inc(&md->event_nr); |
1226 | wake_up(&md->eventq); | |
1227 | } | |
1228 | ||
4e90188b | 1229 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1230 | { |
4e90188b | 1231 | set_capacity(md->disk, size); |
1da177e4 | 1232 | |
1b1dcc1b | 1233 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1234 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1235 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1236 | } |
1237 | ||
1238 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1239 | { | |
165125e1 | 1240 | struct request_queue *q = md->queue; |
1da177e4 LT |
1241 | sector_t size; |
1242 | ||
1243 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1244 | |
1245 | /* | |
1246 | * Wipe any geometry if the size of the table changed. | |
1247 | */ | |
1248 | if (size != get_capacity(md->disk)) | |
1249 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1250 | ||
bfa152fa JN |
1251 | if (md->suspended_bdev) |
1252 | __set_size(md, size); | |
d5816876 MP |
1253 | |
1254 | if (!size) { | |
1255 | dm_table_destroy(t); | |
1da177e4 | 1256 | return 0; |
d5816876 | 1257 | } |
1da177e4 | 1258 | |
2ca3310e AK |
1259 | dm_table_event_callback(t, event_callback, md); |
1260 | ||
1da177e4 LT |
1261 | write_lock(&md->map_lock); |
1262 | md->map = t; | |
2ca3310e | 1263 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1264 | write_unlock(&md->map_lock); |
1265 | ||
1da177e4 LT |
1266 | return 0; |
1267 | } | |
1268 | ||
1269 | static void __unbind(struct mapped_device *md) | |
1270 | { | |
1271 | struct dm_table *map = md->map; | |
1272 | ||
1273 | if (!map) | |
1274 | return; | |
1275 | ||
1276 | dm_table_event_callback(map, NULL, NULL); | |
1277 | write_lock(&md->map_lock); | |
1278 | md->map = NULL; | |
1279 | write_unlock(&md->map_lock); | |
d5816876 | 1280 | dm_table_destroy(map); |
1da177e4 LT |
1281 | } |
1282 | ||
1283 | /* | |
1284 | * Constructor for a new device. | |
1285 | */ | |
2b06cfff | 1286 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1287 | { |
1288 | struct mapped_device *md; | |
1289 | ||
2b06cfff | 1290 | md = alloc_dev(minor); |
1da177e4 LT |
1291 | if (!md) |
1292 | return -ENXIO; | |
1293 | ||
784aae73 MB |
1294 | dm_sysfs_init(md); |
1295 | ||
1da177e4 LT |
1296 | *result = md; |
1297 | return 0; | |
1298 | } | |
1299 | ||
637842cf | 1300 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1301 | { |
1302 | struct mapped_device *md; | |
1da177e4 LT |
1303 | unsigned minor = MINOR(dev); |
1304 | ||
1305 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1306 | return NULL; | |
1307 | ||
f32c10b0 | 1308 | spin_lock(&_minor_lock); |
1da177e4 LT |
1309 | |
1310 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1311 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1312 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1313 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1314 | md = NULL; |
fba9f90e JM |
1315 | goto out; |
1316 | } | |
1da177e4 | 1317 | |
fba9f90e | 1318 | out: |
f32c10b0 | 1319 | spin_unlock(&_minor_lock); |
1da177e4 | 1320 | |
637842cf DT |
1321 | return md; |
1322 | } | |
1323 | ||
d229a958 DT |
1324 | struct mapped_device *dm_get_md(dev_t dev) |
1325 | { | |
1326 | struct mapped_device *md = dm_find_md(dev); | |
1327 | ||
1328 | if (md) | |
1329 | dm_get(md); | |
1330 | ||
1331 | return md; | |
1332 | } | |
1333 | ||
9ade92a9 | 1334 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1335 | { |
9ade92a9 | 1336 | return md->interface_ptr; |
1da177e4 LT |
1337 | } |
1338 | ||
1339 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1340 | { | |
1341 | md->interface_ptr = ptr; | |
1342 | } | |
1343 | ||
1344 | void dm_get(struct mapped_device *md) | |
1345 | { | |
1346 | atomic_inc(&md->holders); | |
1347 | } | |
1348 | ||
72d94861 AK |
1349 | const char *dm_device_name(struct mapped_device *md) |
1350 | { | |
1351 | return md->name; | |
1352 | } | |
1353 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1354 | ||
1da177e4 LT |
1355 | void dm_put(struct mapped_device *md) |
1356 | { | |
1134e5ae | 1357 | struct dm_table *map; |
1da177e4 | 1358 | |
fba9f90e JM |
1359 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1360 | ||
f32c10b0 | 1361 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1362 | map = dm_get_table(md); |
f331c029 TH |
1363 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1364 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1365 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1366 | spin_unlock(&_minor_lock); |
cf222b37 | 1367 | if (!dm_suspended(md)) { |
1da177e4 LT |
1368 | dm_table_presuspend_targets(map); |
1369 | dm_table_postsuspend_targets(map); | |
1370 | } | |
784aae73 | 1371 | dm_sysfs_exit(md); |
1134e5ae | 1372 | dm_table_put(map); |
a1b51e98 | 1373 | __unbind(md); |
1da177e4 LT |
1374 | free_dev(md); |
1375 | } | |
1da177e4 | 1376 | } |
79eb885c | 1377 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1378 | |
46125c1c MB |
1379 | static int dm_wait_for_completion(struct mapped_device *md) |
1380 | { | |
1381 | int r = 0; | |
1382 | ||
1383 | while (1) { | |
1384 | set_current_state(TASK_INTERRUPTIBLE); | |
1385 | ||
1386 | smp_mb(); | |
1387 | if (!atomic_read(&md->pending)) | |
1388 | break; | |
1389 | ||
1390 | if (signal_pending(current)) { | |
1391 | r = -EINTR; | |
1392 | break; | |
1393 | } | |
1394 | ||
1395 | io_schedule(); | |
1396 | } | |
1397 | set_current_state(TASK_RUNNING); | |
1398 | ||
1399 | return r; | |
1400 | } | |
1401 | ||
1da177e4 LT |
1402 | /* |
1403 | * Process the deferred bios | |
1404 | */ | |
6d6f10df | 1405 | static void __flush_deferred_io(struct mapped_device *md) |
1da177e4 | 1406 | { |
6d6f10df | 1407 | struct bio *c; |
1da177e4 | 1408 | |
6d6f10df | 1409 | while ((c = bio_list_pop(&md->deferred))) { |
9e4e5f87 MB |
1410 | if (__split_bio(md, c)) |
1411 | bio_io_error(c); | |
1da177e4 | 1412 | } |
73d410c0 MB |
1413 | |
1414 | clear_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 LT |
1415 | } |
1416 | ||
6d6f10df MB |
1417 | static void __merge_pushback_list(struct mapped_device *md) |
1418 | { | |
1419 | unsigned long flags; | |
1420 | ||
1421 | spin_lock_irqsave(&md->pushback_lock, flags); | |
1422 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1423 | bio_list_merge_head(&md->deferred, &md->pushback); | |
1424 | bio_list_init(&md->pushback); | |
1425 | spin_unlock_irqrestore(&md->pushback_lock, flags); | |
1426 | } | |
1427 | ||
304f3f6a MB |
1428 | static void dm_wq_work(struct work_struct *work) |
1429 | { | |
1430 | struct dm_wq_req *req = container_of(work, struct dm_wq_req, work); | |
1431 | struct mapped_device *md = req->md; | |
1432 | ||
1433 | down_write(&md->io_lock); | |
14377396 | 1434 | __flush_deferred_io(md); |
304f3f6a MB |
1435 | up_write(&md->io_lock); |
1436 | } | |
1437 | ||
14377396 | 1438 | static void dm_wq_queue(struct mapped_device *md, void *context, |
304f3f6a MB |
1439 | struct dm_wq_req *req) |
1440 | { | |
304f3f6a MB |
1441 | req->md = md; |
1442 | req->context = context; | |
1443 | INIT_WORK(&req->work, dm_wq_work); | |
1444 | queue_work(md->wq, &req->work); | |
1445 | } | |
1446 | ||
14377396 | 1447 | static void dm_queue_flush(struct mapped_device *md, void *context) |
304f3f6a MB |
1448 | { |
1449 | struct dm_wq_req req; | |
1450 | ||
14377396 | 1451 | dm_wq_queue(md, context, &req); |
304f3f6a MB |
1452 | flush_workqueue(md->wq); |
1453 | } | |
1454 | ||
1da177e4 LT |
1455 | /* |
1456 | * Swap in a new table (destroying old one). | |
1457 | */ | |
1458 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1459 | { | |
93c534ae | 1460 | int r = -EINVAL; |
1da177e4 | 1461 | |
e61290a4 | 1462 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1463 | |
1464 | /* device must be suspended */ | |
cf222b37 | 1465 | if (!dm_suspended(md)) |
93c534ae | 1466 | goto out; |
1da177e4 | 1467 | |
bfa152fa JN |
1468 | /* without bdev, the device size cannot be changed */ |
1469 | if (!md->suspended_bdev) | |
1470 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1471 | goto out; | |
1472 | ||
1da177e4 LT |
1473 | __unbind(md); |
1474 | r = __bind(md, table); | |
1da177e4 | 1475 | |
93c534ae | 1476 | out: |
e61290a4 | 1477 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1478 | return r; |
1da177e4 LT |
1479 | } |
1480 | ||
1481 | /* | |
1482 | * Functions to lock and unlock any filesystem running on the | |
1483 | * device. | |
1484 | */ | |
2ca3310e | 1485 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1486 | { |
e39e2e95 | 1487 | int r; |
1da177e4 LT |
1488 | |
1489 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1490 | |
e39e2e95 | 1491 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1492 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1493 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1494 | md->frozen_sb = NULL; |
1495 | return r; | |
dfbe03f6 AK |
1496 | } |
1497 | ||
aa8d7c2f AK |
1498 | set_bit(DMF_FROZEN, &md->flags); |
1499 | ||
1da177e4 | 1500 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1501 | * to go away while it is locked. |
1da177e4 LT |
1502 | */ |
1503 | return 0; | |
1504 | } | |
1505 | ||
2ca3310e | 1506 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1507 | { |
aa8d7c2f AK |
1508 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1509 | return; | |
1510 | ||
e39e2e95 | 1511 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1512 | md->frozen_sb = NULL; |
aa8d7c2f | 1513 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1514 | } |
1515 | ||
1516 | /* | |
1517 | * We need to be able to change a mapping table under a mounted | |
1518 | * filesystem. For example we might want to move some data in | |
1519 | * the background. Before the table can be swapped with | |
1520 | * dm_bind_table, dm_suspend must be called to flush any in | |
1521 | * flight bios and ensure that any further io gets deferred. | |
1522 | */ | |
a3d77d35 | 1523 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1524 | { |
2ca3310e | 1525 | struct dm_table *map = NULL; |
1da177e4 | 1526 | DECLARE_WAITQUEUE(wait, current); |
46125c1c | 1527 | int r = 0; |
a3d77d35 | 1528 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1529 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1530 | |
e61290a4 | 1531 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1532 | |
73d410c0 MB |
1533 | if (dm_suspended(md)) { |
1534 | r = -EINVAL; | |
d287483d | 1535 | goto out_unlock; |
73d410c0 | 1536 | } |
1da177e4 LT |
1537 | |
1538 | map = dm_get_table(md); | |
1da177e4 | 1539 | |
2e93ccc1 KU |
1540 | /* |
1541 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1542 | * This flag is cleared before dm_suspend returns. | |
1543 | */ | |
1544 | if (noflush) | |
1545 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1546 | ||
cf222b37 AK |
1547 | /* This does not get reverted if there's an error later. */ |
1548 | dm_table_presuspend_targets(map); | |
1549 | ||
bfa152fa JN |
1550 | /* bdget() can stall if the pending I/Os are not flushed */ |
1551 | if (!noflush) { | |
1552 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1553 | if (!md->suspended_bdev) { | |
1554 | DMWARN("bdget failed in dm_suspend"); | |
1555 | r = -ENOMEM; | |
f431d966 | 1556 | goto out; |
bfa152fa | 1557 | } |
e39e2e95 | 1558 | |
6d6f10df MB |
1559 | /* |
1560 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1561 | * because lock_fs() needs to flush I/Os. | |
1562 | */ | |
1563 | if (do_lockfs) { | |
1564 | r = lock_fs(md); | |
1565 | if (r) | |
1566 | goto out; | |
1567 | } | |
aa8d7c2f | 1568 | } |
1da177e4 LT |
1569 | |
1570 | /* | |
354e0071 | 1571 | * First we set the BLOCK_IO flag so no more ios will be mapped. |
1da177e4 | 1572 | */ |
2ca3310e AK |
1573 | down_write(&md->io_lock); |
1574 | set_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 | 1575 | |
1da177e4 | 1576 | add_wait_queue(&md->wait, &wait); |
2ca3310e | 1577 | up_write(&md->io_lock); |
1da177e4 LT |
1578 | |
1579 | /* unplug */ | |
2ca3310e | 1580 | if (map) |
1da177e4 | 1581 | dm_table_unplug_all(map); |
1da177e4 LT |
1582 | |
1583 | /* | |
46125c1c | 1584 | * Wait for the already-mapped ios to complete. |
1da177e4 | 1585 | */ |
46125c1c | 1586 | r = dm_wait_for_completion(md); |
1da177e4 | 1587 | |
2ca3310e | 1588 | down_write(&md->io_lock); |
1da177e4 LT |
1589 | remove_wait_queue(&md->wait, &wait); |
1590 | ||
6d6f10df MB |
1591 | if (noflush) |
1592 | __merge_pushback_list(md); | |
94d6351e | 1593 | up_write(&md->io_lock); |
2e93ccc1 | 1594 | |
1da177e4 | 1595 | /* were we interrupted ? */ |
46125c1c | 1596 | if (r < 0) { |
14377396 | 1597 | dm_queue_flush(md, NULL); |
73d410c0 | 1598 | |
2ca3310e | 1599 | unlock_fs(md); |
2e93ccc1 | 1600 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1601 | } |
1da177e4 | 1602 | |
cf222b37 | 1603 | dm_table_postsuspend_targets(map); |
1da177e4 | 1604 | |
2ca3310e | 1605 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1606 | |
2ca3310e | 1607 | out: |
e39e2e95 AK |
1608 | if (r && md->suspended_bdev) { |
1609 | bdput(md->suspended_bdev); | |
1610 | md->suspended_bdev = NULL; | |
1611 | } | |
1612 | ||
2ca3310e | 1613 | dm_table_put(map); |
d287483d AK |
1614 | |
1615 | out_unlock: | |
e61290a4 | 1616 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1617 | return r; |
1da177e4 LT |
1618 | } |
1619 | ||
1620 | int dm_resume(struct mapped_device *md) | |
1621 | { | |
cf222b37 | 1622 | int r = -EINVAL; |
cf222b37 | 1623 | struct dm_table *map = NULL; |
1da177e4 | 1624 | |
e61290a4 | 1625 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1626 | if (!dm_suspended(md)) |
cf222b37 | 1627 | goto out; |
cf222b37 AK |
1628 | |
1629 | map = dm_get_table(md); | |
2ca3310e | 1630 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1631 | goto out; |
1da177e4 | 1632 | |
8757b776 MB |
1633 | r = dm_table_resume_targets(map); |
1634 | if (r) | |
1635 | goto out; | |
2ca3310e | 1636 | |
14377396 | 1637 | dm_queue_flush(md, NULL); |
2ca3310e AK |
1638 | |
1639 | unlock_fs(md); | |
1640 | ||
bfa152fa JN |
1641 | if (md->suspended_bdev) { |
1642 | bdput(md->suspended_bdev); | |
1643 | md->suspended_bdev = NULL; | |
1644 | } | |
e39e2e95 | 1645 | |
2ca3310e AK |
1646 | clear_bit(DMF_SUSPENDED, &md->flags); |
1647 | ||
1da177e4 | 1648 | dm_table_unplug_all(map); |
1da177e4 | 1649 | |
69267a30 | 1650 | dm_kobject_uevent(md); |
8560ed6f | 1651 | |
cf222b37 | 1652 | r = 0; |
2ca3310e | 1653 | |
cf222b37 AK |
1654 | out: |
1655 | dm_table_put(map); | |
e61290a4 | 1656 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1657 | |
cf222b37 | 1658 | return r; |
1da177e4 LT |
1659 | } |
1660 | ||
1661 | /*----------------------------------------------------------------- | |
1662 | * Event notification. | |
1663 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1664 | void dm_kobject_uevent(struct mapped_device *md) |
1665 | { | |
ed9e1982 | 1666 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1667 | } |
1668 | ||
7a8c3d3b MA |
1669 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1670 | { | |
1671 | return atomic_add_return(1, &md->uevent_seq); | |
1672 | } | |
1673 | ||
1da177e4 LT |
1674 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1675 | { | |
1676 | return atomic_read(&md->event_nr); | |
1677 | } | |
1678 | ||
1679 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1680 | { | |
1681 | return wait_event_interruptible(md->eventq, | |
1682 | (event_nr != atomic_read(&md->event_nr))); | |
1683 | } | |
1684 | ||
7a8c3d3b MA |
1685 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1686 | { | |
1687 | unsigned long flags; | |
1688 | ||
1689 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1690 | list_add(elist, &md->uevent_list); | |
1691 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1692 | } | |
1693 | ||
1da177e4 LT |
1694 | /* |
1695 | * The gendisk is only valid as long as you have a reference | |
1696 | * count on 'md'. | |
1697 | */ | |
1698 | struct gendisk *dm_disk(struct mapped_device *md) | |
1699 | { | |
1700 | return md->disk; | |
1701 | } | |
1702 | ||
784aae73 MB |
1703 | struct kobject *dm_kobject(struct mapped_device *md) |
1704 | { | |
1705 | return &md->kobj; | |
1706 | } | |
1707 | ||
1708 | /* | |
1709 | * struct mapped_device should not be exported outside of dm.c | |
1710 | * so use this check to verify that kobj is part of md structure | |
1711 | */ | |
1712 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1713 | { | |
1714 | struct mapped_device *md; | |
1715 | ||
1716 | md = container_of(kobj, struct mapped_device, kobj); | |
1717 | if (&md->kobj != kobj) | |
1718 | return NULL; | |
1719 | ||
1720 | dm_get(md); | |
1721 | return md; | |
1722 | } | |
1723 | ||
1da177e4 LT |
1724 | int dm_suspended(struct mapped_device *md) |
1725 | { | |
1726 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1727 | } | |
1728 | ||
2e93ccc1 KU |
1729 | int dm_noflush_suspending(struct dm_target *ti) |
1730 | { | |
1731 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1732 | int r = __noflush_suspending(md); | |
1733 | ||
1734 | dm_put(md); | |
1735 | ||
1736 | return r; | |
1737 | } | |
1738 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1739 | ||
1da177e4 LT |
1740 | static struct block_device_operations dm_blk_dops = { |
1741 | .open = dm_blk_open, | |
1742 | .release = dm_blk_close, | |
aa129a22 | 1743 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1744 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1745 | .owner = THIS_MODULE |
1746 | }; | |
1747 | ||
1748 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1749 | ||
1750 | /* | |
1751 | * module hooks | |
1752 | */ | |
1753 | module_init(dm_init); | |
1754 | module_exit(dm_exit); | |
1755 | ||
1756 | module_param(major, uint, 0); | |
1757 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1758 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1759 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1760 | MODULE_LICENSE("GPL"); |