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 { | |
103 | enum { | |
304f3f6a MB |
104 | DM_WQ_FLUSH_DEFERRED, |
105 | } type; | |
106 | struct work_struct work; | |
107 | struct mapped_device *md; | |
108 | void *context; | |
109 | }; | |
110 | ||
1da177e4 | 111 | struct mapped_device { |
2ca3310e | 112 | struct rw_semaphore io_lock; |
e61290a4 | 113 | struct mutex suspend_lock; |
2e93ccc1 | 114 | spinlock_t pushback_lock; |
1da177e4 LT |
115 | rwlock_t map_lock; |
116 | atomic_t holders; | |
5c6bd75d | 117 | atomic_t open_count; |
1da177e4 LT |
118 | |
119 | unsigned long flags; | |
120 | ||
165125e1 | 121 | struct request_queue *queue; |
1da177e4 | 122 | struct gendisk *disk; |
7e51f257 | 123 | char name[16]; |
1da177e4 LT |
124 | |
125 | void *interface_ptr; | |
126 | ||
127 | /* | |
128 | * A list of ios that arrived while we were suspended. | |
129 | */ | |
130 | atomic_t pending; | |
131 | wait_queue_head_t wait; | |
74859364 | 132 | struct bio_list deferred; |
2e93ccc1 | 133 | struct bio_list pushback; |
1da177e4 | 134 | |
304f3f6a MB |
135 | /* |
136 | * Processing queue (flush/barriers) | |
137 | */ | |
138 | struct workqueue_struct *wq; | |
139 | ||
1da177e4 LT |
140 | /* |
141 | * The current mapping. | |
142 | */ | |
143 | struct dm_table *map; | |
144 | ||
145 | /* | |
146 | * io objects are allocated from here. | |
147 | */ | |
148 | mempool_t *io_pool; | |
149 | mempool_t *tio_pool; | |
150 | ||
9faf400f SB |
151 | struct bio_set *bs; |
152 | ||
1da177e4 LT |
153 | /* |
154 | * Event handling. | |
155 | */ | |
156 | atomic_t event_nr; | |
157 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
158 | atomic_t uevent_seq; |
159 | struct list_head uevent_list; | |
160 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
161 | |
162 | /* | |
163 | * freeze/thaw support require holding onto a super block | |
164 | */ | |
165 | struct super_block *frozen_sb; | |
e39e2e95 | 166 | struct block_device *suspended_bdev; |
3ac51e74 DW |
167 | |
168 | /* forced geometry settings */ | |
169 | struct hd_geometry geometry; | |
784aae73 MB |
170 | |
171 | /* sysfs handle */ | |
172 | struct kobject kobj; | |
1da177e4 LT |
173 | }; |
174 | ||
175 | #define MIN_IOS 256 | |
e18b890b CL |
176 | static struct kmem_cache *_io_cache; |
177 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
178 | static struct kmem_cache *_rq_tio_cache; |
179 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 180 | |
1da177e4 LT |
181 | static int __init local_init(void) |
182 | { | |
51157b4a | 183 | int r = -ENOMEM; |
1da177e4 | 184 | |
1da177e4 | 185 | /* allocate a slab for the dm_ios */ |
028867ac | 186 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 187 | if (!_io_cache) |
51157b4a | 188 | return r; |
1da177e4 LT |
189 | |
190 | /* allocate a slab for the target ios */ | |
028867ac | 191 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
192 | if (!_tio_cache) |
193 | goto out_free_io_cache; | |
1da177e4 | 194 | |
8fbf26ad KU |
195 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
196 | if (!_rq_tio_cache) | |
197 | goto out_free_tio_cache; | |
198 | ||
199 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
200 | if (!_rq_bio_info_cache) | |
201 | goto out_free_rq_tio_cache; | |
202 | ||
51e5b2bd | 203 | r = dm_uevent_init(); |
51157b4a | 204 | if (r) |
8fbf26ad | 205 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 206 | |
1da177e4 LT |
207 | _major = major; |
208 | r = register_blkdev(_major, _name); | |
51157b4a KU |
209 | if (r < 0) |
210 | goto out_uevent_exit; | |
1da177e4 LT |
211 | |
212 | if (!_major) | |
213 | _major = r; | |
214 | ||
215 | return 0; | |
51157b4a KU |
216 | |
217 | out_uevent_exit: | |
218 | dm_uevent_exit(); | |
8fbf26ad KU |
219 | out_free_rq_bio_info_cache: |
220 | kmem_cache_destroy(_rq_bio_info_cache); | |
221 | out_free_rq_tio_cache: | |
222 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
223 | out_free_tio_cache: |
224 | kmem_cache_destroy(_tio_cache); | |
225 | out_free_io_cache: | |
226 | kmem_cache_destroy(_io_cache); | |
227 | ||
228 | return r; | |
1da177e4 LT |
229 | } |
230 | ||
231 | static void local_exit(void) | |
232 | { | |
8fbf26ad KU |
233 | kmem_cache_destroy(_rq_bio_info_cache); |
234 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
235 | kmem_cache_destroy(_tio_cache); |
236 | kmem_cache_destroy(_io_cache); | |
00d59405 | 237 | unregister_blkdev(_major, _name); |
51e5b2bd | 238 | dm_uevent_exit(); |
1da177e4 LT |
239 | |
240 | _major = 0; | |
241 | ||
242 | DMINFO("cleaned up"); | |
243 | } | |
244 | ||
b9249e55 | 245 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
246 | local_init, |
247 | dm_target_init, | |
248 | dm_linear_init, | |
249 | dm_stripe_init, | |
945fa4d2 | 250 | dm_kcopyd_init, |
1da177e4 LT |
251 | dm_interface_init, |
252 | }; | |
253 | ||
b9249e55 | 254 | static void (*_exits[])(void) = { |
1da177e4 LT |
255 | local_exit, |
256 | dm_target_exit, | |
257 | dm_linear_exit, | |
258 | dm_stripe_exit, | |
945fa4d2 | 259 | dm_kcopyd_exit, |
1da177e4 LT |
260 | dm_interface_exit, |
261 | }; | |
262 | ||
263 | static int __init dm_init(void) | |
264 | { | |
265 | const int count = ARRAY_SIZE(_inits); | |
266 | ||
267 | int r, i; | |
268 | ||
269 | for (i = 0; i < count; i++) { | |
270 | r = _inits[i](); | |
271 | if (r) | |
272 | goto bad; | |
273 | } | |
274 | ||
275 | return 0; | |
276 | ||
277 | bad: | |
278 | while (i--) | |
279 | _exits[i](); | |
280 | ||
281 | return r; | |
282 | } | |
283 | ||
284 | static void __exit dm_exit(void) | |
285 | { | |
286 | int i = ARRAY_SIZE(_exits); | |
287 | ||
288 | while (i--) | |
289 | _exits[i](); | |
290 | } | |
291 | ||
292 | /* | |
293 | * Block device functions | |
294 | */ | |
fe5f9f2c | 295 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
296 | { |
297 | struct mapped_device *md; | |
298 | ||
fba9f90e JM |
299 | spin_lock(&_minor_lock); |
300 | ||
fe5f9f2c | 301 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
302 | if (!md) |
303 | goto out; | |
304 | ||
5c6bd75d AK |
305 | if (test_bit(DMF_FREEING, &md->flags) || |
306 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
307 | md = NULL; |
308 | goto out; | |
309 | } | |
310 | ||
1da177e4 | 311 | dm_get(md); |
5c6bd75d | 312 | atomic_inc(&md->open_count); |
fba9f90e JM |
313 | |
314 | out: | |
315 | spin_unlock(&_minor_lock); | |
316 | ||
317 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
318 | } |
319 | ||
fe5f9f2c | 320 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 321 | { |
fe5f9f2c | 322 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 323 | atomic_dec(&md->open_count); |
1da177e4 LT |
324 | dm_put(md); |
325 | return 0; | |
326 | } | |
327 | ||
5c6bd75d AK |
328 | int dm_open_count(struct mapped_device *md) |
329 | { | |
330 | return atomic_read(&md->open_count); | |
331 | } | |
332 | ||
333 | /* | |
334 | * Guarantees nothing is using the device before it's deleted. | |
335 | */ | |
336 | int dm_lock_for_deletion(struct mapped_device *md) | |
337 | { | |
338 | int r = 0; | |
339 | ||
340 | spin_lock(&_minor_lock); | |
341 | ||
342 | if (dm_open_count(md)) | |
343 | r = -EBUSY; | |
344 | else | |
345 | set_bit(DMF_DELETING, &md->flags); | |
346 | ||
347 | spin_unlock(&_minor_lock); | |
348 | ||
349 | return r; | |
350 | } | |
351 | ||
3ac51e74 DW |
352 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
353 | { | |
354 | struct mapped_device *md = bdev->bd_disk->private_data; | |
355 | ||
356 | return dm_get_geometry(md, geo); | |
357 | } | |
358 | ||
fe5f9f2c | 359 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
360 | unsigned int cmd, unsigned long arg) |
361 | { | |
fe5f9f2c AV |
362 | struct mapped_device *md = bdev->bd_disk->private_data; |
363 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
364 | struct dm_target *tgt; |
365 | int r = -ENOTTY; | |
366 | ||
aa129a22 MB |
367 | if (!map || !dm_table_get_size(map)) |
368 | goto out; | |
369 | ||
370 | /* We only support devices that have a single target */ | |
371 | if (dm_table_get_num_targets(map) != 1) | |
372 | goto out; | |
373 | ||
374 | tgt = dm_table_get_target(map, 0); | |
375 | ||
376 | if (dm_suspended(md)) { | |
377 | r = -EAGAIN; | |
378 | goto out; | |
379 | } | |
380 | ||
381 | if (tgt->type->ioctl) | |
647b3d00 | 382 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
383 | |
384 | out: | |
385 | dm_table_put(map); | |
386 | ||
aa129a22 MB |
387 | return r; |
388 | } | |
389 | ||
028867ac | 390 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
391 | { |
392 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
393 | } | |
394 | ||
028867ac | 395 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
396 | { |
397 | mempool_free(io, md->io_pool); | |
398 | } | |
399 | ||
028867ac | 400 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
401 | { |
402 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
403 | } | |
404 | ||
028867ac | 405 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
406 | { |
407 | mempool_free(tio, md->tio_pool); | |
408 | } | |
409 | ||
3eaf840e JNN |
410 | static void start_io_acct(struct dm_io *io) |
411 | { | |
412 | struct mapped_device *md = io->md; | |
c9959059 | 413 | int cpu; |
3eaf840e JNN |
414 | |
415 | io->start_time = jiffies; | |
416 | ||
074a7aca TH |
417 | cpu = part_stat_lock(); |
418 | part_round_stats(cpu, &dm_disk(md)->part0); | |
419 | part_stat_unlock(); | |
420 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
421 | } |
422 | ||
d221d2e7 | 423 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
424 | { |
425 | struct mapped_device *md = io->md; | |
426 | struct bio *bio = io->bio; | |
427 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 428 | int pending, cpu; |
3eaf840e JNN |
429 | int rw = bio_data_dir(bio); |
430 | ||
074a7aca TH |
431 | cpu = part_stat_lock(); |
432 | part_round_stats(cpu, &dm_disk(md)->part0); | |
433 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
434 | part_stat_unlock(); | |
3eaf840e | 435 | |
074a7aca TH |
436 | dm_disk(md)->part0.in_flight = pending = |
437 | atomic_dec_return(&md->pending); | |
3eaf840e | 438 | |
d221d2e7 MP |
439 | /* nudge anyone waiting on suspend queue */ |
440 | if (!pending) | |
441 | wake_up(&md->wait); | |
3eaf840e JNN |
442 | } |
443 | ||
1da177e4 LT |
444 | /* |
445 | * Add the bio to the list of deferred io. | |
446 | */ | |
447 | static int queue_io(struct mapped_device *md, struct bio *bio) | |
448 | { | |
2ca3310e | 449 | down_write(&md->io_lock); |
1da177e4 LT |
450 | |
451 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 452 | up_write(&md->io_lock); |
1da177e4 LT |
453 | return 1; |
454 | } | |
455 | ||
456 | bio_list_add(&md->deferred, bio); | |
457 | ||
2ca3310e | 458 | up_write(&md->io_lock); |
1da177e4 LT |
459 | return 0; /* deferred successfully */ |
460 | } | |
461 | ||
462 | /* | |
463 | * Everyone (including functions in this file), should use this | |
464 | * function to access the md->map field, and make sure they call | |
465 | * dm_table_put() when finished. | |
466 | */ | |
467 | struct dm_table *dm_get_table(struct mapped_device *md) | |
468 | { | |
469 | struct dm_table *t; | |
470 | ||
471 | read_lock(&md->map_lock); | |
472 | t = md->map; | |
473 | if (t) | |
474 | dm_table_get(t); | |
475 | read_unlock(&md->map_lock); | |
476 | ||
477 | return t; | |
478 | } | |
479 | ||
3ac51e74 DW |
480 | /* |
481 | * Get the geometry associated with a dm device | |
482 | */ | |
483 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
484 | { | |
485 | *geo = md->geometry; | |
486 | ||
487 | return 0; | |
488 | } | |
489 | ||
490 | /* | |
491 | * Set the geometry of a device. | |
492 | */ | |
493 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
494 | { | |
495 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
496 | ||
497 | if (geo->start > sz) { | |
498 | DMWARN("Start sector is beyond the geometry limits."); | |
499 | return -EINVAL; | |
500 | } | |
501 | ||
502 | md->geometry = *geo; | |
503 | ||
504 | return 0; | |
505 | } | |
506 | ||
1da177e4 LT |
507 | /*----------------------------------------------------------------- |
508 | * CRUD START: | |
509 | * A more elegant soln is in the works that uses the queue | |
510 | * merge fn, unfortunately there are a couple of changes to | |
511 | * the block layer that I want to make for this. So in the | |
512 | * interests of getting something for people to use I give | |
513 | * you this clearly demarcated crap. | |
514 | *---------------------------------------------------------------*/ | |
515 | ||
2e93ccc1 KU |
516 | static int __noflush_suspending(struct mapped_device *md) |
517 | { | |
518 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
519 | } | |
520 | ||
1da177e4 LT |
521 | /* |
522 | * Decrements the number of outstanding ios that a bio has been | |
523 | * cloned into, completing the original io if necc. | |
524 | */ | |
858119e1 | 525 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 526 | { |
2e93ccc1 KU |
527 | unsigned long flags; |
528 | ||
529 | /* Push-back supersedes any I/O errors */ | |
530 | if (error && !(io->error > 0 && __noflush_suspending(io->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 | */ | |
540 | spin_lock_irqsave(&io->md->pushback_lock, flags); | |
541 | if (__noflush_suspending(io->md)) | |
542 | bio_list_add(&io->md->pushback, io->bio); | |
543 | else | |
544 | /* noflush suspend was interrupted. */ | |
545 | io->error = -EIO; | |
546 | spin_unlock_irqrestore(&io->md->pushback_lock, flags); | |
547 | } | |
548 | ||
d221d2e7 | 549 | end_io_acct(io); |
1da177e4 | 550 | |
2e93ccc1 | 551 | if (io->error != DM_ENDIO_REQUEUE) { |
5f3ea37c | 552 | trace_block_bio_complete(io->md->queue, io->bio); |
2e93ccc1 | 553 | |
6712ecf8 | 554 | bio_endio(io->bio, io->error); |
2e93ccc1 | 555 | } |
2056a782 | 556 | |
1da177e4 LT |
557 | free_io(io->md, io); |
558 | } | |
559 | } | |
560 | ||
6712ecf8 | 561 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
562 | { |
563 | int r = 0; | |
028867ac | 564 | struct dm_target_io *tio = bio->bi_private; |
9faf400f | 565 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
566 | dm_endio_fn endio = tio->ti->type->end_io; |
567 | ||
1da177e4 LT |
568 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
569 | error = -EIO; | |
570 | ||
571 | if (endio) { | |
572 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
573 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
574 | /* | |
575 | * error and requeue request are handled | |
576 | * in dec_pending(). | |
577 | */ | |
1da177e4 | 578 | error = r; |
45cbcd79 KU |
579 | else if (r == DM_ENDIO_INCOMPLETE) |
580 | /* The target will handle the io */ | |
6712ecf8 | 581 | return; |
45cbcd79 KU |
582 | else if (r) { |
583 | DMWARN("unimplemented target endio return value: %d", r); | |
584 | BUG(); | |
585 | } | |
1da177e4 LT |
586 | } |
587 | ||
9faf400f SB |
588 | dec_pending(tio->io, error); |
589 | ||
590 | /* | |
591 | * Store md for cleanup instead of tio which is about to get freed. | |
592 | */ | |
593 | bio->bi_private = md->bs; | |
594 | ||
1da177e4 | 595 | bio_put(bio); |
9faf400f | 596 | free_tio(md, tio); |
1da177e4 LT |
597 | } |
598 | ||
599 | static sector_t max_io_len(struct mapped_device *md, | |
600 | sector_t sector, struct dm_target *ti) | |
601 | { | |
602 | sector_t offset = sector - ti->begin; | |
603 | sector_t len = ti->len - offset; | |
604 | ||
605 | /* | |
606 | * Does the target need to split even further ? | |
607 | */ | |
608 | if (ti->split_io) { | |
609 | sector_t boundary; | |
610 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
611 | - offset; | |
612 | if (len > boundary) | |
613 | len = boundary; | |
614 | } | |
615 | ||
616 | return len; | |
617 | } | |
618 | ||
619 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 620 | struct dm_target_io *tio) |
1da177e4 LT |
621 | { |
622 | int r; | |
2056a782 | 623 | sector_t sector; |
9faf400f | 624 | struct mapped_device *md; |
1da177e4 LT |
625 | |
626 | /* | |
627 | * Sanity checks. | |
628 | */ | |
629 | BUG_ON(!clone->bi_size); | |
630 | ||
631 | clone->bi_end_io = clone_endio; | |
632 | clone->bi_private = tio; | |
633 | ||
634 | /* | |
635 | * Map the clone. If r == 0 we don't need to do | |
636 | * anything, the target has assumed ownership of | |
637 | * this io. | |
638 | */ | |
639 | atomic_inc(&tio->io->io_count); | |
2056a782 | 640 | sector = clone->bi_sector; |
1da177e4 | 641 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 642 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 643 | /* the bio has been remapped so dispatch it */ |
2056a782 | 644 | |
5f3ea37c | 645 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
646 | tio->io->bio->bi_bdev->bd_dev, |
647 | clone->bi_sector, sector); | |
2056a782 | 648 | |
1da177e4 | 649 | generic_make_request(clone); |
2e93ccc1 KU |
650 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
651 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
652 | md = tio->io->md; |
653 | dec_pending(tio->io, r); | |
654 | /* | |
655 | * Store bio_set for cleanup. | |
656 | */ | |
657 | clone->bi_private = md->bs; | |
1da177e4 | 658 | bio_put(clone); |
9faf400f | 659 | free_tio(md, tio); |
45cbcd79 KU |
660 | } else if (r) { |
661 | DMWARN("unimplemented target map return value: %d", r); | |
662 | BUG(); | |
1da177e4 LT |
663 | } |
664 | } | |
665 | ||
666 | struct clone_info { | |
667 | struct mapped_device *md; | |
668 | struct dm_table *map; | |
669 | struct bio *bio; | |
670 | struct dm_io *io; | |
671 | sector_t sector; | |
672 | sector_t sector_count; | |
673 | unsigned short idx; | |
674 | }; | |
675 | ||
3676347a PO |
676 | static void dm_bio_destructor(struct bio *bio) |
677 | { | |
9faf400f SB |
678 | struct bio_set *bs = bio->bi_private; |
679 | ||
680 | bio_free(bio, bs); | |
3676347a PO |
681 | } |
682 | ||
1da177e4 LT |
683 | /* |
684 | * Creates a little bio that is just does part of a bvec. | |
685 | */ | |
686 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
687 | unsigned short idx, unsigned int offset, | |
9faf400f | 688 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
689 | { |
690 | struct bio *clone; | |
691 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
692 | ||
9faf400f | 693 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 694 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
695 | *clone->bi_io_vec = *bv; |
696 | ||
697 | clone->bi_sector = sector; | |
698 | clone->bi_bdev = bio->bi_bdev; | |
699 | clone->bi_rw = bio->bi_rw; | |
700 | clone->bi_vcnt = 1; | |
701 | clone->bi_size = to_bytes(len); | |
702 | clone->bi_io_vec->bv_offset = offset; | |
703 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 704 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 LT |
705 | |
706 | return clone; | |
707 | } | |
708 | ||
709 | /* | |
710 | * Creates a bio that consists of range of complete bvecs. | |
711 | */ | |
712 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
713 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 714 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
715 | { |
716 | struct bio *clone; | |
717 | ||
9faf400f SB |
718 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
719 | __bio_clone(clone, bio); | |
720 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
721 | clone->bi_sector = sector; |
722 | clone->bi_idx = idx; | |
723 | clone->bi_vcnt = idx + bv_count; | |
724 | clone->bi_size = to_bytes(len); | |
725 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
726 | ||
727 | return clone; | |
728 | } | |
729 | ||
512875bd | 730 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
731 | { |
732 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
733 | struct dm_target *ti; |
734 | sector_t len = 0, max; | |
028867ac | 735 | struct dm_target_io *tio; |
1da177e4 | 736 | |
512875bd JN |
737 | ti = dm_table_find_target(ci->map, ci->sector); |
738 | if (!dm_target_is_valid(ti)) | |
739 | return -EIO; | |
740 | ||
741 | max = max_io_len(ci->md, ci->sector, ti); | |
742 | ||
1da177e4 LT |
743 | /* |
744 | * Allocate a target io object. | |
745 | */ | |
746 | tio = alloc_tio(ci->md); | |
747 | tio->io = ci->io; | |
748 | tio->ti = ti; | |
749 | memset(&tio->info, 0, sizeof(tio->info)); | |
750 | ||
751 | if (ci->sector_count <= max) { | |
752 | /* | |
753 | * Optimise for the simple case where we can do all of | |
754 | * the remaining io with a single clone. | |
755 | */ | |
756 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
757 | bio->bi_vcnt - ci->idx, ci->sector_count, |
758 | ci->md->bs); | |
1da177e4 LT |
759 | __map_bio(ti, clone, tio); |
760 | ci->sector_count = 0; | |
761 | ||
762 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
763 | /* | |
764 | * There are some bvecs that don't span targets. | |
765 | * Do as many of these as possible. | |
766 | */ | |
767 | int i; | |
768 | sector_t remaining = max; | |
769 | sector_t bv_len; | |
770 | ||
771 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
772 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
773 | ||
774 | if (bv_len > remaining) | |
775 | break; | |
776 | ||
777 | remaining -= bv_len; | |
778 | len += bv_len; | |
779 | } | |
780 | ||
9faf400f SB |
781 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
782 | ci->md->bs); | |
1da177e4 LT |
783 | __map_bio(ti, clone, tio); |
784 | ||
785 | ci->sector += len; | |
786 | ci->sector_count -= len; | |
787 | ci->idx = i; | |
788 | ||
789 | } else { | |
790 | /* | |
d2044a94 | 791 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
792 | */ |
793 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
794 | sector_t remaining = to_sector(bv->bv_len); |
795 | unsigned int offset = 0; | |
1da177e4 | 796 | |
d2044a94 AK |
797 | do { |
798 | if (offset) { | |
799 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
800 | if (!dm_target_is_valid(ti)) |
801 | return -EIO; | |
802 | ||
d2044a94 | 803 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 804 | |
d2044a94 AK |
805 | tio = alloc_tio(ci->md); |
806 | tio->io = ci->io; | |
807 | tio->ti = ti; | |
808 | memset(&tio->info, 0, sizeof(tio->info)); | |
809 | } | |
810 | ||
811 | len = min(remaining, max); | |
812 | ||
813 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
814 | bv->bv_offset + offset, len, |
815 | ci->md->bs); | |
d2044a94 AK |
816 | |
817 | __map_bio(ti, clone, tio); | |
818 | ||
819 | ci->sector += len; | |
820 | ci->sector_count -= len; | |
821 | offset += to_bytes(len); | |
822 | } while (remaining -= len); | |
1da177e4 | 823 | |
1da177e4 LT |
824 | ci->idx++; |
825 | } | |
512875bd JN |
826 | |
827 | return 0; | |
1da177e4 LT |
828 | } |
829 | ||
830 | /* | |
831 | * Split the bio into several clones. | |
832 | */ | |
9e4e5f87 | 833 | static int __split_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
834 | { |
835 | struct clone_info ci; | |
512875bd | 836 | int error = 0; |
1da177e4 LT |
837 | |
838 | ci.map = dm_get_table(md); | |
9e4e5f87 MB |
839 | if (unlikely(!ci.map)) |
840 | return -EIO; | |
ab4c1424 AK |
841 | if (unlikely(bio_barrier(bio) && !dm_table_barrier_ok(ci.map))) { |
842 | dm_table_put(ci.map); | |
843 | bio_endio(bio, -EOPNOTSUPP); | |
844 | return 0; | |
845 | } | |
1da177e4 LT |
846 | ci.md = md; |
847 | ci.bio = bio; | |
848 | ci.io = alloc_io(md); | |
849 | ci.io->error = 0; | |
850 | atomic_set(&ci.io->io_count, 1); | |
851 | ci.io->bio = bio; | |
852 | ci.io->md = md; | |
853 | ci.sector = bio->bi_sector; | |
854 | ci.sector_count = bio_sectors(bio); | |
855 | ci.idx = bio->bi_idx; | |
856 | ||
3eaf840e | 857 | start_io_acct(ci.io); |
512875bd JN |
858 | while (ci.sector_count && !error) |
859 | error = __clone_and_map(&ci); | |
1da177e4 LT |
860 | |
861 | /* drop the extra reference count */ | |
512875bd | 862 | dec_pending(ci.io, error); |
1da177e4 | 863 | dm_table_put(ci.map); |
9e4e5f87 MB |
864 | |
865 | return 0; | |
1da177e4 LT |
866 | } |
867 | /*----------------------------------------------------------------- | |
868 | * CRUD END | |
869 | *---------------------------------------------------------------*/ | |
870 | ||
f6fccb12 MB |
871 | static int dm_merge_bvec(struct request_queue *q, |
872 | struct bvec_merge_data *bvm, | |
873 | struct bio_vec *biovec) | |
874 | { | |
875 | struct mapped_device *md = q->queuedata; | |
876 | struct dm_table *map = dm_get_table(md); | |
877 | struct dm_target *ti; | |
878 | sector_t max_sectors; | |
5037108a | 879 | int max_size = 0; |
f6fccb12 MB |
880 | |
881 | if (unlikely(!map)) | |
5037108a | 882 | goto out; |
f6fccb12 MB |
883 | |
884 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
885 | if (!dm_target_is_valid(ti)) |
886 | goto out_table; | |
f6fccb12 MB |
887 | |
888 | /* | |
889 | * Find maximum amount of I/O that won't need splitting | |
890 | */ | |
891 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
892 | (sector_t) BIO_MAX_SECTORS); | |
893 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
894 | if (max_size < 0) | |
895 | max_size = 0; | |
896 | ||
897 | /* | |
898 | * merge_bvec_fn() returns number of bytes | |
899 | * it can accept at this offset | |
900 | * max is precomputed maximal io size | |
901 | */ | |
902 | if (max_size && ti->type->merge) | |
903 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
904 | ||
b01cd5ac | 905 | out_table: |
5037108a MP |
906 | dm_table_put(map); |
907 | ||
908 | out: | |
f6fccb12 MB |
909 | /* |
910 | * Always allow an entire first page | |
911 | */ | |
912 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
913 | max_size = biovec->bv_len; | |
914 | ||
f6fccb12 MB |
915 | return max_size; |
916 | } | |
917 | ||
1da177e4 LT |
918 | /* |
919 | * The request function that just remaps the bio built up by | |
920 | * dm_merge_bvec. | |
921 | */ | |
165125e1 | 922 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 923 | { |
9e4e5f87 | 924 | int r = -EIO; |
12f03a49 | 925 | int rw = bio_data_dir(bio); |
1da177e4 | 926 | struct mapped_device *md = q->queuedata; |
c9959059 | 927 | int cpu; |
1da177e4 | 928 | |
2ca3310e | 929 | down_read(&md->io_lock); |
1da177e4 | 930 | |
074a7aca TH |
931 | cpu = part_stat_lock(); |
932 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
933 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
934 | part_stat_unlock(); | |
12f03a49 | 935 | |
1da177e4 LT |
936 | /* |
937 | * If we're suspended we have to queue | |
938 | * this io for later. | |
939 | */ | |
940 | while (test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 941 | up_read(&md->io_lock); |
1da177e4 | 942 | |
9e4e5f87 MB |
943 | if (bio_rw(bio) != READA) |
944 | r = queue_io(md, bio); | |
1da177e4 | 945 | |
9e4e5f87 MB |
946 | if (r <= 0) |
947 | goto out_req; | |
1da177e4 LT |
948 | |
949 | /* | |
950 | * We're in a while loop, because someone could suspend | |
951 | * before we get to the following read lock. | |
952 | */ | |
2ca3310e | 953 | down_read(&md->io_lock); |
1da177e4 LT |
954 | } |
955 | ||
9e4e5f87 | 956 | r = __split_bio(md, bio); |
2ca3310e | 957 | up_read(&md->io_lock); |
9e4e5f87 MB |
958 | |
959 | out_req: | |
960 | if (r < 0) | |
961 | bio_io_error(bio); | |
962 | ||
1da177e4 LT |
963 | return 0; |
964 | } | |
965 | ||
165125e1 | 966 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
967 | { |
968 | struct mapped_device *md = q->queuedata; | |
969 | struct dm_table *map = dm_get_table(md); | |
970 | ||
971 | if (map) { | |
972 | dm_table_unplug_all(map); | |
973 | dm_table_put(map); | |
974 | } | |
975 | } | |
976 | ||
977 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
978 | { | |
8a57dfc6 CS |
979 | int r = bdi_bits; |
980 | struct mapped_device *md = congested_data; | |
981 | struct dm_table *map; | |
1da177e4 | 982 | |
8a57dfc6 CS |
983 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { |
984 | map = dm_get_table(md); | |
985 | if (map) { | |
986 | r = dm_table_any_congested(map, bdi_bits); | |
987 | dm_table_put(map); | |
988 | } | |
989 | } | |
990 | ||
1da177e4 LT |
991 | return r; |
992 | } | |
993 | ||
994 | /*----------------------------------------------------------------- | |
995 | * An IDR is used to keep track of allocated minor numbers. | |
996 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
997 | static DEFINE_IDR(_minor_idr); |
998 | ||
2b06cfff | 999 | static void free_minor(int minor) |
1da177e4 | 1000 | { |
f32c10b0 | 1001 | spin_lock(&_minor_lock); |
1da177e4 | 1002 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1003 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1004 | } |
1005 | ||
1006 | /* | |
1007 | * See if the device with a specific minor # is free. | |
1008 | */ | |
cf13ab8e | 1009 | static int specific_minor(int minor) |
1da177e4 LT |
1010 | { |
1011 | int r, m; | |
1012 | ||
1013 | if (minor >= (1 << MINORBITS)) | |
1014 | return -EINVAL; | |
1015 | ||
62f75c2f JM |
1016 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1017 | if (!r) | |
1018 | return -ENOMEM; | |
1019 | ||
f32c10b0 | 1020 | spin_lock(&_minor_lock); |
1da177e4 LT |
1021 | |
1022 | if (idr_find(&_minor_idr, minor)) { | |
1023 | r = -EBUSY; | |
1024 | goto out; | |
1025 | } | |
1026 | ||
ba61fdd1 | 1027 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1028 | if (r) |
1da177e4 | 1029 | goto out; |
1da177e4 LT |
1030 | |
1031 | if (m != minor) { | |
1032 | idr_remove(&_minor_idr, m); | |
1033 | r = -EBUSY; | |
1034 | goto out; | |
1035 | } | |
1036 | ||
1037 | out: | |
f32c10b0 | 1038 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1039 | return r; |
1040 | } | |
1041 | ||
cf13ab8e | 1042 | static int next_free_minor(int *minor) |
1da177e4 | 1043 | { |
2b06cfff | 1044 | int r, m; |
1da177e4 | 1045 | |
1da177e4 | 1046 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1047 | if (!r) |
1048 | return -ENOMEM; | |
1049 | ||
f32c10b0 | 1050 | spin_lock(&_minor_lock); |
1da177e4 | 1051 | |
ba61fdd1 | 1052 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1053 | if (r) |
1da177e4 | 1054 | goto out; |
1da177e4 LT |
1055 | |
1056 | if (m >= (1 << MINORBITS)) { | |
1057 | idr_remove(&_minor_idr, m); | |
1058 | r = -ENOSPC; | |
1059 | goto out; | |
1060 | } | |
1061 | ||
1062 | *minor = m; | |
1063 | ||
1064 | out: | |
f32c10b0 | 1065 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1066 | return r; |
1067 | } | |
1068 | ||
1069 | static struct block_device_operations dm_blk_dops; | |
1070 | ||
1071 | /* | |
1072 | * Allocate and initialise a blank device with a given minor. | |
1073 | */ | |
2b06cfff | 1074 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1075 | { |
1076 | int r; | |
cf13ab8e | 1077 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1078 | void *old_md; |
1da177e4 LT |
1079 | |
1080 | if (!md) { | |
1081 | DMWARN("unable to allocate device, out of memory."); | |
1082 | return NULL; | |
1083 | } | |
1084 | ||
10da4f79 | 1085 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1086 | goto bad_module_get; |
10da4f79 | 1087 | |
1da177e4 | 1088 | /* get a minor number for the dev */ |
2b06cfff | 1089 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1090 | r = next_free_minor(&minor); |
2b06cfff | 1091 | else |
cf13ab8e | 1092 | r = specific_minor(minor); |
1da177e4 | 1093 | if (r < 0) |
6ed7ade8 | 1094 | goto bad_minor; |
1da177e4 | 1095 | |
2ca3310e | 1096 | init_rwsem(&md->io_lock); |
e61290a4 | 1097 | mutex_init(&md->suspend_lock); |
2e93ccc1 | 1098 | spin_lock_init(&md->pushback_lock); |
1da177e4 LT |
1099 | rwlock_init(&md->map_lock); |
1100 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1101 | atomic_set(&md->open_count, 0); |
1da177e4 | 1102 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1103 | atomic_set(&md->uevent_seq, 0); |
1104 | INIT_LIST_HEAD(&md->uevent_list); | |
1105 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1106 | |
1107 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1108 | if (!md->queue) | |
6ed7ade8 | 1109 | goto bad_queue; |
1da177e4 LT |
1110 | |
1111 | md->queue->queuedata = md; | |
1112 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1113 | md->queue->backing_dev_info.congested_data = md; | |
1114 | blk_queue_make_request(md->queue, dm_request); | |
daef265f | 1115 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1116 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1117 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1118 | |
93d2341c | 1119 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1120 | if (!md->io_pool) |
6ed7ade8 | 1121 | goto bad_io_pool; |
1da177e4 | 1122 | |
93d2341c | 1123 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1124 | if (!md->tio_pool) |
6ed7ade8 | 1125 | goto bad_tio_pool; |
1da177e4 | 1126 | |
bb799ca0 | 1127 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1128 | if (!md->bs) |
1129 | goto bad_no_bioset; | |
1130 | ||
1da177e4 LT |
1131 | md->disk = alloc_disk(1); |
1132 | if (!md->disk) | |
6ed7ade8 | 1133 | goto bad_disk; |
1da177e4 | 1134 | |
f0b04115 JM |
1135 | atomic_set(&md->pending, 0); |
1136 | init_waitqueue_head(&md->wait); | |
1137 | init_waitqueue_head(&md->eventq); | |
1138 | ||
1da177e4 LT |
1139 | md->disk->major = _major; |
1140 | md->disk->first_minor = minor; | |
1141 | md->disk->fops = &dm_blk_dops; | |
1142 | md->disk->queue = md->queue; | |
1143 | md->disk->private_data = md; | |
1144 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1145 | add_disk(md->disk); | |
7e51f257 | 1146 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1147 | |
304f3f6a MB |
1148 | md->wq = create_singlethread_workqueue("kdmflush"); |
1149 | if (!md->wq) | |
1150 | goto bad_thread; | |
1151 | ||
ba61fdd1 | 1152 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1153 | spin_lock(&_minor_lock); |
ba61fdd1 | 1154 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1155 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1156 | |
1157 | BUG_ON(old_md != MINOR_ALLOCED); | |
1158 | ||
1da177e4 LT |
1159 | return md; |
1160 | ||
304f3f6a MB |
1161 | bad_thread: |
1162 | put_disk(md->disk); | |
6ed7ade8 | 1163 | bad_disk: |
9faf400f | 1164 | bioset_free(md->bs); |
6ed7ade8 | 1165 | bad_no_bioset: |
1da177e4 | 1166 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1167 | bad_tio_pool: |
1da177e4 | 1168 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1169 | bad_io_pool: |
1312f40e | 1170 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1171 | bad_queue: |
1da177e4 | 1172 | free_minor(minor); |
6ed7ade8 | 1173 | bad_minor: |
10da4f79 | 1174 | module_put(THIS_MODULE); |
6ed7ade8 | 1175 | bad_module_get: |
1da177e4 LT |
1176 | kfree(md); |
1177 | return NULL; | |
1178 | } | |
1179 | ||
ae9da83f JN |
1180 | static void unlock_fs(struct mapped_device *md); |
1181 | ||
1da177e4 LT |
1182 | static void free_dev(struct mapped_device *md) |
1183 | { | |
f331c029 | 1184 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1185 | |
d9dde59b | 1186 | if (md->suspended_bdev) { |
ae9da83f | 1187 | unlock_fs(md); |
d9dde59b JN |
1188 | bdput(md->suspended_bdev); |
1189 | } | |
304f3f6a | 1190 | destroy_workqueue(md->wq); |
1da177e4 LT |
1191 | mempool_destroy(md->tio_pool); |
1192 | mempool_destroy(md->io_pool); | |
9faf400f | 1193 | bioset_free(md->bs); |
1da177e4 | 1194 | del_gendisk(md->disk); |
63d94e48 | 1195 | free_minor(minor); |
fba9f90e JM |
1196 | |
1197 | spin_lock(&_minor_lock); | |
1198 | md->disk->private_data = NULL; | |
1199 | spin_unlock(&_minor_lock); | |
1200 | ||
1da177e4 | 1201 | put_disk(md->disk); |
1312f40e | 1202 | blk_cleanup_queue(md->queue); |
10da4f79 | 1203 | module_put(THIS_MODULE); |
1da177e4 LT |
1204 | kfree(md); |
1205 | } | |
1206 | ||
1207 | /* | |
1208 | * Bind a table to the device. | |
1209 | */ | |
1210 | static void event_callback(void *context) | |
1211 | { | |
7a8c3d3b MA |
1212 | unsigned long flags; |
1213 | LIST_HEAD(uevents); | |
1da177e4 LT |
1214 | struct mapped_device *md = (struct mapped_device *) context; |
1215 | ||
7a8c3d3b MA |
1216 | spin_lock_irqsave(&md->uevent_lock, flags); |
1217 | list_splice_init(&md->uevent_list, &uevents); | |
1218 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1219 | ||
ed9e1982 | 1220 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1221 | |
1da177e4 LT |
1222 | atomic_inc(&md->event_nr); |
1223 | wake_up(&md->eventq); | |
1224 | } | |
1225 | ||
4e90188b | 1226 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1227 | { |
4e90188b | 1228 | set_capacity(md->disk, size); |
1da177e4 | 1229 | |
1b1dcc1b | 1230 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1231 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1232 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1233 | } |
1234 | ||
1235 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1236 | { | |
165125e1 | 1237 | struct request_queue *q = md->queue; |
1da177e4 LT |
1238 | sector_t size; |
1239 | ||
1240 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1241 | |
1242 | /* | |
1243 | * Wipe any geometry if the size of the table changed. | |
1244 | */ | |
1245 | if (size != get_capacity(md->disk)) | |
1246 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1247 | ||
bfa152fa JN |
1248 | if (md->suspended_bdev) |
1249 | __set_size(md, size); | |
d5816876 MP |
1250 | |
1251 | if (!size) { | |
1252 | dm_table_destroy(t); | |
1da177e4 | 1253 | return 0; |
d5816876 | 1254 | } |
1da177e4 | 1255 | |
2ca3310e AK |
1256 | dm_table_event_callback(t, event_callback, md); |
1257 | ||
1da177e4 LT |
1258 | write_lock(&md->map_lock); |
1259 | md->map = t; | |
2ca3310e | 1260 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1261 | write_unlock(&md->map_lock); |
1262 | ||
1da177e4 LT |
1263 | return 0; |
1264 | } | |
1265 | ||
1266 | static void __unbind(struct mapped_device *md) | |
1267 | { | |
1268 | struct dm_table *map = md->map; | |
1269 | ||
1270 | if (!map) | |
1271 | return; | |
1272 | ||
1273 | dm_table_event_callback(map, NULL, NULL); | |
1274 | write_lock(&md->map_lock); | |
1275 | md->map = NULL; | |
1276 | write_unlock(&md->map_lock); | |
d5816876 | 1277 | dm_table_destroy(map); |
1da177e4 LT |
1278 | } |
1279 | ||
1280 | /* | |
1281 | * Constructor for a new device. | |
1282 | */ | |
2b06cfff | 1283 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1284 | { |
1285 | struct mapped_device *md; | |
1286 | ||
2b06cfff | 1287 | md = alloc_dev(minor); |
1da177e4 LT |
1288 | if (!md) |
1289 | return -ENXIO; | |
1290 | ||
784aae73 MB |
1291 | dm_sysfs_init(md); |
1292 | ||
1da177e4 LT |
1293 | *result = md; |
1294 | return 0; | |
1295 | } | |
1296 | ||
637842cf | 1297 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1298 | { |
1299 | struct mapped_device *md; | |
1da177e4 LT |
1300 | unsigned minor = MINOR(dev); |
1301 | ||
1302 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1303 | return NULL; | |
1304 | ||
f32c10b0 | 1305 | spin_lock(&_minor_lock); |
1da177e4 LT |
1306 | |
1307 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1308 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1309 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1310 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1311 | md = NULL; |
fba9f90e JM |
1312 | goto out; |
1313 | } | |
1da177e4 | 1314 | |
fba9f90e | 1315 | out: |
f32c10b0 | 1316 | spin_unlock(&_minor_lock); |
1da177e4 | 1317 | |
637842cf DT |
1318 | return md; |
1319 | } | |
1320 | ||
d229a958 DT |
1321 | struct mapped_device *dm_get_md(dev_t dev) |
1322 | { | |
1323 | struct mapped_device *md = dm_find_md(dev); | |
1324 | ||
1325 | if (md) | |
1326 | dm_get(md); | |
1327 | ||
1328 | return md; | |
1329 | } | |
1330 | ||
9ade92a9 | 1331 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1332 | { |
9ade92a9 | 1333 | return md->interface_ptr; |
1da177e4 LT |
1334 | } |
1335 | ||
1336 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1337 | { | |
1338 | md->interface_ptr = ptr; | |
1339 | } | |
1340 | ||
1341 | void dm_get(struct mapped_device *md) | |
1342 | { | |
1343 | atomic_inc(&md->holders); | |
1344 | } | |
1345 | ||
72d94861 AK |
1346 | const char *dm_device_name(struct mapped_device *md) |
1347 | { | |
1348 | return md->name; | |
1349 | } | |
1350 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1351 | ||
1da177e4 LT |
1352 | void dm_put(struct mapped_device *md) |
1353 | { | |
1134e5ae | 1354 | struct dm_table *map; |
1da177e4 | 1355 | |
fba9f90e JM |
1356 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1357 | ||
f32c10b0 | 1358 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1359 | map = dm_get_table(md); |
f331c029 TH |
1360 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1361 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1362 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1363 | spin_unlock(&_minor_lock); |
cf222b37 | 1364 | if (!dm_suspended(md)) { |
1da177e4 LT |
1365 | dm_table_presuspend_targets(map); |
1366 | dm_table_postsuspend_targets(map); | |
1367 | } | |
784aae73 | 1368 | dm_sysfs_exit(md); |
1134e5ae | 1369 | dm_table_put(map); |
a1b51e98 | 1370 | __unbind(md); |
1da177e4 LT |
1371 | free_dev(md); |
1372 | } | |
1da177e4 | 1373 | } |
79eb885c | 1374 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1375 | |
46125c1c MB |
1376 | static int dm_wait_for_completion(struct mapped_device *md) |
1377 | { | |
1378 | int r = 0; | |
1379 | ||
1380 | while (1) { | |
1381 | set_current_state(TASK_INTERRUPTIBLE); | |
1382 | ||
1383 | smp_mb(); | |
1384 | if (!atomic_read(&md->pending)) | |
1385 | break; | |
1386 | ||
1387 | if (signal_pending(current)) { | |
1388 | r = -EINTR; | |
1389 | break; | |
1390 | } | |
1391 | ||
1392 | io_schedule(); | |
1393 | } | |
1394 | set_current_state(TASK_RUNNING); | |
1395 | ||
1396 | return r; | |
1397 | } | |
1398 | ||
1da177e4 LT |
1399 | /* |
1400 | * Process the deferred bios | |
1401 | */ | |
6d6f10df | 1402 | static void __flush_deferred_io(struct mapped_device *md) |
1da177e4 | 1403 | { |
6d6f10df | 1404 | struct bio *c; |
1da177e4 | 1405 | |
6d6f10df | 1406 | while ((c = bio_list_pop(&md->deferred))) { |
9e4e5f87 MB |
1407 | if (__split_bio(md, c)) |
1408 | bio_io_error(c); | |
1da177e4 | 1409 | } |
73d410c0 MB |
1410 | |
1411 | clear_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 LT |
1412 | } |
1413 | ||
6d6f10df MB |
1414 | static void __merge_pushback_list(struct mapped_device *md) |
1415 | { | |
1416 | unsigned long flags; | |
1417 | ||
1418 | spin_lock_irqsave(&md->pushback_lock, flags); | |
1419 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1420 | bio_list_merge_head(&md->deferred, &md->pushback); | |
1421 | bio_list_init(&md->pushback); | |
1422 | spin_unlock_irqrestore(&md->pushback_lock, flags); | |
1423 | } | |
1424 | ||
304f3f6a MB |
1425 | static void dm_wq_work(struct work_struct *work) |
1426 | { | |
1427 | struct dm_wq_req *req = container_of(work, struct dm_wq_req, work); | |
1428 | struct mapped_device *md = req->md; | |
1429 | ||
1430 | down_write(&md->io_lock); | |
1431 | switch (req->type) { | |
304f3f6a MB |
1432 | case DM_WQ_FLUSH_DEFERRED: |
1433 | __flush_deferred_io(md); | |
1434 | break; | |
1435 | default: | |
1436 | DMERR("dm_wq_work: unrecognised work type %d", req->type); | |
1437 | BUG(); | |
1438 | } | |
1439 | up_write(&md->io_lock); | |
1440 | } | |
1441 | ||
1442 | static void dm_wq_queue(struct mapped_device *md, int type, void *context, | |
1443 | struct dm_wq_req *req) | |
1444 | { | |
1445 | req->type = type; | |
1446 | req->md = md; | |
1447 | req->context = context; | |
1448 | INIT_WORK(&req->work, dm_wq_work); | |
1449 | queue_work(md->wq, &req->work); | |
1450 | } | |
1451 | ||
1452 | static void dm_queue_flush(struct mapped_device *md, int type, void *context) | |
1453 | { | |
1454 | struct dm_wq_req req; | |
1455 | ||
1456 | dm_wq_queue(md, type, context, &req); | |
1457 | flush_workqueue(md->wq); | |
1458 | } | |
1459 | ||
1da177e4 LT |
1460 | /* |
1461 | * Swap in a new table (destroying old one). | |
1462 | */ | |
1463 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1464 | { | |
93c534ae | 1465 | int r = -EINVAL; |
1da177e4 | 1466 | |
e61290a4 | 1467 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1468 | |
1469 | /* device must be suspended */ | |
cf222b37 | 1470 | if (!dm_suspended(md)) |
93c534ae | 1471 | goto out; |
1da177e4 | 1472 | |
bfa152fa JN |
1473 | /* without bdev, the device size cannot be changed */ |
1474 | if (!md->suspended_bdev) | |
1475 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1476 | goto out; | |
1477 | ||
1da177e4 LT |
1478 | __unbind(md); |
1479 | r = __bind(md, table); | |
1da177e4 | 1480 | |
93c534ae | 1481 | out: |
e61290a4 | 1482 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1483 | return r; |
1da177e4 LT |
1484 | } |
1485 | ||
1486 | /* | |
1487 | * Functions to lock and unlock any filesystem running on the | |
1488 | * device. | |
1489 | */ | |
2ca3310e | 1490 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1491 | { |
e39e2e95 | 1492 | int r; |
1da177e4 LT |
1493 | |
1494 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1495 | |
e39e2e95 | 1496 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1497 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1498 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1499 | md->frozen_sb = NULL; |
1500 | return r; | |
dfbe03f6 AK |
1501 | } |
1502 | ||
aa8d7c2f AK |
1503 | set_bit(DMF_FROZEN, &md->flags); |
1504 | ||
1da177e4 | 1505 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1506 | * to go away while it is locked. |
1da177e4 LT |
1507 | */ |
1508 | return 0; | |
1509 | } | |
1510 | ||
2ca3310e | 1511 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1512 | { |
aa8d7c2f AK |
1513 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1514 | return; | |
1515 | ||
e39e2e95 | 1516 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1517 | md->frozen_sb = NULL; |
aa8d7c2f | 1518 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1519 | } |
1520 | ||
1521 | /* | |
1522 | * We need to be able to change a mapping table under a mounted | |
1523 | * filesystem. For example we might want to move some data in | |
1524 | * the background. Before the table can be swapped with | |
1525 | * dm_bind_table, dm_suspend must be called to flush any in | |
1526 | * flight bios and ensure that any further io gets deferred. | |
1527 | */ | |
a3d77d35 | 1528 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1529 | { |
2ca3310e | 1530 | struct dm_table *map = NULL; |
1da177e4 | 1531 | DECLARE_WAITQUEUE(wait, current); |
46125c1c | 1532 | int r = 0; |
a3d77d35 | 1533 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1534 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1535 | |
e61290a4 | 1536 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1537 | |
73d410c0 MB |
1538 | if (dm_suspended(md)) { |
1539 | r = -EINVAL; | |
d287483d | 1540 | goto out_unlock; |
73d410c0 | 1541 | } |
1da177e4 LT |
1542 | |
1543 | map = dm_get_table(md); | |
1da177e4 | 1544 | |
2e93ccc1 KU |
1545 | /* |
1546 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1547 | * This flag is cleared before dm_suspend returns. | |
1548 | */ | |
1549 | if (noflush) | |
1550 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1551 | ||
cf222b37 AK |
1552 | /* This does not get reverted if there's an error later. */ |
1553 | dm_table_presuspend_targets(map); | |
1554 | ||
bfa152fa JN |
1555 | /* bdget() can stall if the pending I/Os are not flushed */ |
1556 | if (!noflush) { | |
1557 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1558 | if (!md->suspended_bdev) { | |
1559 | DMWARN("bdget failed in dm_suspend"); | |
1560 | r = -ENOMEM; | |
f431d966 | 1561 | goto out; |
bfa152fa | 1562 | } |
e39e2e95 | 1563 | |
6d6f10df MB |
1564 | /* |
1565 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1566 | * because lock_fs() needs to flush I/Os. | |
1567 | */ | |
1568 | if (do_lockfs) { | |
1569 | r = lock_fs(md); | |
1570 | if (r) | |
1571 | goto out; | |
1572 | } | |
aa8d7c2f | 1573 | } |
1da177e4 LT |
1574 | |
1575 | /* | |
354e0071 | 1576 | * First we set the BLOCK_IO flag so no more ios will be mapped. |
1da177e4 | 1577 | */ |
2ca3310e AK |
1578 | down_write(&md->io_lock); |
1579 | set_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 | 1580 | |
1da177e4 | 1581 | add_wait_queue(&md->wait, &wait); |
2ca3310e | 1582 | up_write(&md->io_lock); |
1da177e4 LT |
1583 | |
1584 | /* unplug */ | |
2ca3310e | 1585 | if (map) |
1da177e4 | 1586 | dm_table_unplug_all(map); |
1da177e4 LT |
1587 | |
1588 | /* | |
46125c1c | 1589 | * Wait for the already-mapped ios to complete. |
1da177e4 | 1590 | */ |
46125c1c | 1591 | r = dm_wait_for_completion(md); |
1da177e4 | 1592 | |
2ca3310e | 1593 | down_write(&md->io_lock); |
1da177e4 LT |
1594 | remove_wait_queue(&md->wait, &wait); |
1595 | ||
6d6f10df MB |
1596 | if (noflush) |
1597 | __merge_pushback_list(md); | |
94d6351e | 1598 | up_write(&md->io_lock); |
2e93ccc1 | 1599 | |
1da177e4 | 1600 | /* were we interrupted ? */ |
46125c1c | 1601 | if (r < 0) { |
304f3f6a | 1602 | dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL); |
73d410c0 | 1603 | |
2ca3310e | 1604 | unlock_fs(md); |
2e93ccc1 | 1605 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1606 | } |
1da177e4 | 1607 | |
cf222b37 | 1608 | dm_table_postsuspend_targets(map); |
1da177e4 | 1609 | |
2ca3310e | 1610 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1611 | |
2ca3310e | 1612 | out: |
e39e2e95 AK |
1613 | if (r && md->suspended_bdev) { |
1614 | bdput(md->suspended_bdev); | |
1615 | md->suspended_bdev = NULL; | |
1616 | } | |
1617 | ||
2ca3310e | 1618 | dm_table_put(map); |
d287483d AK |
1619 | |
1620 | out_unlock: | |
e61290a4 | 1621 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1622 | return r; |
1da177e4 LT |
1623 | } |
1624 | ||
1625 | int dm_resume(struct mapped_device *md) | |
1626 | { | |
cf222b37 | 1627 | int r = -EINVAL; |
cf222b37 | 1628 | struct dm_table *map = NULL; |
1da177e4 | 1629 | |
e61290a4 | 1630 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1631 | if (!dm_suspended(md)) |
cf222b37 | 1632 | goto out; |
cf222b37 AK |
1633 | |
1634 | map = dm_get_table(md); | |
2ca3310e | 1635 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1636 | goto out; |
1da177e4 | 1637 | |
8757b776 MB |
1638 | r = dm_table_resume_targets(map); |
1639 | if (r) | |
1640 | goto out; | |
2ca3310e | 1641 | |
304f3f6a | 1642 | dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL); |
2ca3310e AK |
1643 | |
1644 | unlock_fs(md); | |
1645 | ||
bfa152fa JN |
1646 | if (md->suspended_bdev) { |
1647 | bdput(md->suspended_bdev); | |
1648 | md->suspended_bdev = NULL; | |
1649 | } | |
e39e2e95 | 1650 | |
2ca3310e AK |
1651 | clear_bit(DMF_SUSPENDED, &md->flags); |
1652 | ||
1da177e4 | 1653 | dm_table_unplug_all(map); |
1da177e4 | 1654 | |
69267a30 | 1655 | dm_kobject_uevent(md); |
8560ed6f | 1656 | |
cf222b37 | 1657 | r = 0; |
2ca3310e | 1658 | |
cf222b37 AK |
1659 | out: |
1660 | dm_table_put(map); | |
e61290a4 | 1661 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1662 | |
cf222b37 | 1663 | return r; |
1da177e4 LT |
1664 | } |
1665 | ||
1666 | /*----------------------------------------------------------------- | |
1667 | * Event notification. | |
1668 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1669 | void dm_kobject_uevent(struct mapped_device *md) |
1670 | { | |
ed9e1982 | 1671 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1672 | } |
1673 | ||
7a8c3d3b MA |
1674 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1675 | { | |
1676 | return atomic_add_return(1, &md->uevent_seq); | |
1677 | } | |
1678 | ||
1da177e4 LT |
1679 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1680 | { | |
1681 | return atomic_read(&md->event_nr); | |
1682 | } | |
1683 | ||
1684 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1685 | { | |
1686 | return wait_event_interruptible(md->eventq, | |
1687 | (event_nr != atomic_read(&md->event_nr))); | |
1688 | } | |
1689 | ||
7a8c3d3b MA |
1690 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1691 | { | |
1692 | unsigned long flags; | |
1693 | ||
1694 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1695 | list_add(elist, &md->uevent_list); | |
1696 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1697 | } | |
1698 | ||
1da177e4 LT |
1699 | /* |
1700 | * The gendisk is only valid as long as you have a reference | |
1701 | * count on 'md'. | |
1702 | */ | |
1703 | struct gendisk *dm_disk(struct mapped_device *md) | |
1704 | { | |
1705 | return md->disk; | |
1706 | } | |
1707 | ||
784aae73 MB |
1708 | struct kobject *dm_kobject(struct mapped_device *md) |
1709 | { | |
1710 | return &md->kobj; | |
1711 | } | |
1712 | ||
1713 | /* | |
1714 | * struct mapped_device should not be exported outside of dm.c | |
1715 | * so use this check to verify that kobj is part of md structure | |
1716 | */ | |
1717 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1718 | { | |
1719 | struct mapped_device *md; | |
1720 | ||
1721 | md = container_of(kobj, struct mapped_device, kobj); | |
1722 | if (&md->kobj != kobj) | |
1723 | return NULL; | |
1724 | ||
1725 | dm_get(md); | |
1726 | return md; | |
1727 | } | |
1728 | ||
1da177e4 LT |
1729 | int dm_suspended(struct mapped_device *md) |
1730 | { | |
1731 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1732 | } | |
1733 | ||
2e93ccc1 KU |
1734 | int dm_noflush_suspending(struct dm_target *ti) |
1735 | { | |
1736 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1737 | int r = __noflush_suspending(md); | |
1738 | ||
1739 | dm_put(md); | |
1740 | ||
1741 | return r; | |
1742 | } | |
1743 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1744 | ||
1da177e4 LT |
1745 | static struct block_device_operations dm_blk_dops = { |
1746 | .open = dm_blk_open, | |
1747 | .release = dm_blk_close, | |
aa129a22 | 1748 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1749 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1750 | .owner = THIS_MODULE |
1751 | }; | |
1752 | ||
1753 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1754 | ||
1755 | /* | |
1756 | * module hooks | |
1757 | */ | |
1758 | module_init(dm_init); | |
1759 | module_exit(dm_exit); | |
1760 | ||
1761 | module_param(major, uint, 0); | |
1762 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1763 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1764 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1765 | MODULE_LICENSE("GPL"); |