]>
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 | */ | |
1da177e4 | 102 | struct mapped_device { |
2ca3310e | 103 | struct rw_semaphore io_lock; |
e61290a4 | 104 | struct mutex suspend_lock; |
2e93ccc1 | 105 | spinlock_t pushback_lock; |
1da177e4 LT |
106 | rwlock_t map_lock; |
107 | atomic_t holders; | |
5c6bd75d | 108 | atomic_t open_count; |
1da177e4 LT |
109 | |
110 | unsigned long flags; | |
111 | ||
165125e1 | 112 | struct request_queue *queue; |
1da177e4 | 113 | struct gendisk *disk; |
7e51f257 | 114 | char name[16]; |
1da177e4 LT |
115 | |
116 | void *interface_ptr; | |
117 | ||
118 | /* | |
119 | * A list of ios that arrived while we were suspended. | |
120 | */ | |
121 | atomic_t pending; | |
122 | wait_queue_head_t wait; | |
53d5914f | 123 | struct work_struct work; |
74859364 | 124 | struct bio_list deferred; |
2e93ccc1 | 125 | struct bio_list pushback; |
1da177e4 | 126 | |
304f3f6a MB |
127 | /* |
128 | * Processing queue (flush/barriers) | |
129 | */ | |
130 | struct workqueue_struct *wq; | |
131 | ||
1da177e4 LT |
132 | /* |
133 | * The current mapping. | |
134 | */ | |
135 | struct dm_table *map; | |
136 | ||
137 | /* | |
138 | * io objects are allocated from here. | |
139 | */ | |
140 | mempool_t *io_pool; | |
141 | mempool_t *tio_pool; | |
142 | ||
9faf400f SB |
143 | struct bio_set *bs; |
144 | ||
1da177e4 LT |
145 | /* |
146 | * Event handling. | |
147 | */ | |
148 | atomic_t event_nr; | |
149 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
150 | atomic_t uevent_seq; |
151 | struct list_head uevent_list; | |
152 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
153 | |
154 | /* | |
155 | * freeze/thaw support require holding onto a super block | |
156 | */ | |
157 | struct super_block *frozen_sb; | |
e39e2e95 | 158 | struct block_device *suspended_bdev; |
3ac51e74 DW |
159 | |
160 | /* forced geometry settings */ | |
161 | struct hd_geometry geometry; | |
784aae73 MB |
162 | |
163 | /* sysfs handle */ | |
164 | struct kobject kobj; | |
1da177e4 LT |
165 | }; |
166 | ||
167 | #define MIN_IOS 256 | |
e18b890b CL |
168 | static struct kmem_cache *_io_cache; |
169 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
170 | static struct kmem_cache *_rq_tio_cache; |
171 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 172 | |
1da177e4 LT |
173 | static int __init local_init(void) |
174 | { | |
51157b4a | 175 | int r = -ENOMEM; |
1da177e4 | 176 | |
1da177e4 | 177 | /* allocate a slab for the dm_ios */ |
028867ac | 178 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 179 | if (!_io_cache) |
51157b4a | 180 | return r; |
1da177e4 LT |
181 | |
182 | /* allocate a slab for the target ios */ | |
028867ac | 183 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
184 | if (!_tio_cache) |
185 | goto out_free_io_cache; | |
1da177e4 | 186 | |
8fbf26ad KU |
187 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
188 | if (!_rq_tio_cache) | |
189 | goto out_free_tio_cache; | |
190 | ||
191 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
192 | if (!_rq_bio_info_cache) | |
193 | goto out_free_rq_tio_cache; | |
194 | ||
51e5b2bd | 195 | r = dm_uevent_init(); |
51157b4a | 196 | if (r) |
8fbf26ad | 197 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 198 | |
1da177e4 LT |
199 | _major = major; |
200 | r = register_blkdev(_major, _name); | |
51157b4a KU |
201 | if (r < 0) |
202 | goto out_uevent_exit; | |
1da177e4 LT |
203 | |
204 | if (!_major) | |
205 | _major = r; | |
206 | ||
207 | return 0; | |
51157b4a KU |
208 | |
209 | out_uevent_exit: | |
210 | dm_uevent_exit(); | |
8fbf26ad KU |
211 | out_free_rq_bio_info_cache: |
212 | kmem_cache_destroy(_rq_bio_info_cache); | |
213 | out_free_rq_tio_cache: | |
214 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
215 | out_free_tio_cache: |
216 | kmem_cache_destroy(_tio_cache); | |
217 | out_free_io_cache: | |
218 | kmem_cache_destroy(_io_cache); | |
219 | ||
220 | return r; | |
1da177e4 LT |
221 | } |
222 | ||
223 | static void local_exit(void) | |
224 | { | |
8fbf26ad KU |
225 | kmem_cache_destroy(_rq_bio_info_cache); |
226 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
227 | kmem_cache_destroy(_tio_cache); |
228 | kmem_cache_destroy(_io_cache); | |
00d59405 | 229 | unregister_blkdev(_major, _name); |
51e5b2bd | 230 | dm_uevent_exit(); |
1da177e4 LT |
231 | |
232 | _major = 0; | |
233 | ||
234 | DMINFO("cleaned up"); | |
235 | } | |
236 | ||
b9249e55 | 237 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
238 | local_init, |
239 | dm_target_init, | |
240 | dm_linear_init, | |
241 | dm_stripe_init, | |
945fa4d2 | 242 | dm_kcopyd_init, |
1da177e4 LT |
243 | dm_interface_init, |
244 | }; | |
245 | ||
b9249e55 | 246 | static void (*_exits[])(void) = { |
1da177e4 LT |
247 | local_exit, |
248 | dm_target_exit, | |
249 | dm_linear_exit, | |
250 | dm_stripe_exit, | |
945fa4d2 | 251 | dm_kcopyd_exit, |
1da177e4 LT |
252 | dm_interface_exit, |
253 | }; | |
254 | ||
255 | static int __init dm_init(void) | |
256 | { | |
257 | const int count = ARRAY_SIZE(_inits); | |
258 | ||
259 | int r, i; | |
260 | ||
261 | for (i = 0; i < count; i++) { | |
262 | r = _inits[i](); | |
263 | if (r) | |
264 | goto bad; | |
265 | } | |
266 | ||
267 | return 0; | |
268 | ||
269 | bad: | |
270 | while (i--) | |
271 | _exits[i](); | |
272 | ||
273 | return r; | |
274 | } | |
275 | ||
276 | static void __exit dm_exit(void) | |
277 | { | |
278 | int i = ARRAY_SIZE(_exits); | |
279 | ||
280 | while (i--) | |
281 | _exits[i](); | |
282 | } | |
283 | ||
284 | /* | |
285 | * Block device functions | |
286 | */ | |
fe5f9f2c | 287 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
288 | { |
289 | struct mapped_device *md; | |
290 | ||
fba9f90e JM |
291 | spin_lock(&_minor_lock); |
292 | ||
fe5f9f2c | 293 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
294 | if (!md) |
295 | goto out; | |
296 | ||
5c6bd75d AK |
297 | if (test_bit(DMF_FREEING, &md->flags) || |
298 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
299 | md = NULL; |
300 | goto out; | |
301 | } | |
302 | ||
1da177e4 | 303 | dm_get(md); |
5c6bd75d | 304 | atomic_inc(&md->open_count); |
fba9f90e JM |
305 | |
306 | out: | |
307 | spin_unlock(&_minor_lock); | |
308 | ||
309 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
310 | } |
311 | ||
fe5f9f2c | 312 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 313 | { |
fe5f9f2c | 314 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 315 | atomic_dec(&md->open_count); |
1da177e4 LT |
316 | dm_put(md); |
317 | return 0; | |
318 | } | |
319 | ||
5c6bd75d AK |
320 | int dm_open_count(struct mapped_device *md) |
321 | { | |
322 | return atomic_read(&md->open_count); | |
323 | } | |
324 | ||
325 | /* | |
326 | * Guarantees nothing is using the device before it's deleted. | |
327 | */ | |
328 | int dm_lock_for_deletion(struct mapped_device *md) | |
329 | { | |
330 | int r = 0; | |
331 | ||
332 | spin_lock(&_minor_lock); | |
333 | ||
334 | if (dm_open_count(md)) | |
335 | r = -EBUSY; | |
336 | else | |
337 | set_bit(DMF_DELETING, &md->flags); | |
338 | ||
339 | spin_unlock(&_minor_lock); | |
340 | ||
341 | return r; | |
342 | } | |
343 | ||
3ac51e74 DW |
344 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
345 | { | |
346 | struct mapped_device *md = bdev->bd_disk->private_data; | |
347 | ||
348 | return dm_get_geometry(md, geo); | |
349 | } | |
350 | ||
fe5f9f2c | 351 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
352 | unsigned int cmd, unsigned long arg) |
353 | { | |
fe5f9f2c AV |
354 | struct mapped_device *md = bdev->bd_disk->private_data; |
355 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
356 | struct dm_target *tgt; |
357 | int r = -ENOTTY; | |
358 | ||
aa129a22 MB |
359 | if (!map || !dm_table_get_size(map)) |
360 | goto out; | |
361 | ||
362 | /* We only support devices that have a single target */ | |
363 | if (dm_table_get_num_targets(map) != 1) | |
364 | goto out; | |
365 | ||
366 | tgt = dm_table_get_target(map, 0); | |
367 | ||
368 | if (dm_suspended(md)) { | |
369 | r = -EAGAIN; | |
370 | goto out; | |
371 | } | |
372 | ||
373 | if (tgt->type->ioctl) | |
647b3d00 | 374 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
375 | |
376 | out: | |
377 | dm_table_put(map); | |
378 | ||
aa129a22 MB |
379 | return r; |
380 | } | |
381 | ||
028867ac | 382 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
383 | { |
384 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
385 | } | |
386 | ||
028867ac | 387 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
388 | { |
389 | mempool_free(io, md->io_pool); | |
390 | } | |
391 | ||
028867ac | 392 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
393 | { |
394 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
395 | } | |
396 | ||
028867ac | 397 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
398 | { |
399 | mempool_free(tio, md->tio_pool); | |
400 | } | |
401 | ||
3eaf840e JNN |
402 | static void start_io_acct(struct dm_io *io) |
403 | { | |
404 | struct mapped_device *md = io->md; | |
c9959059 | 405 | int cpu; |
3eaf840e JNN |
406 | |
407 | io->start_time = jiffies; | |
408 | ||
074a7aca TH |
409 | cpu = part_stat_lock(); |
410 | part_round_stats(cpu, &dm_disk(md)->part0); | |
411 | part_stat_unlock(); | |
412 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
413 | } |
414 | ||
d221d2e7 | 415 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
416 | { |
417 | struct mapped_device *md = io->md; | |
418 | struct bio *bio = io->bio; | |
419 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 420 | int pending, cpu; |
3eaf840e JNN |
421 | int rw = bio_data_dir(bio); |
422 | ||
074a7aca TH |
423 | cpu = part_stat_lock(); |
424 | part_round_stats(cpu, &dm_disk(md)->part0); | |
425 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
426 | part_stat_unlock(); | |
3eaf840e | 427 | |
074a7aca TH |
428 | dm_disk(md)->part0.in_flight = pending = |
429 | atomic_dec_return(&md->pending); | |
3eaf840e | 430 | |
d221d2e7 MP |
431 | /* nudge anyone waiting on suspend queue */ |
432 | if (!pending) | |
433 | wake_up(&md->wait); | |
3eaf840e JNN |
434 | } |
435 | ||
1da177e4 LT |
436 | /* |
437 | * Add the bio to the list of deferred io. | |
438 | */ | |
439 | static int queue_io(struct mapped_device *md, struct bio *bio) | |
440 | { | |
2ca3310e | 441 | down_write(&md->io_lock); |
1da177e4 LT |
442 | |
443 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 444 | up_write(&md->io_lock); |
1da177e4 LT |
445 | return 1; |
446 | } | |
447 | ||
448 | bio_list_add(&md->deferred, bio); | |
449 | ||
2ca3310e | 450 | up_write(&md->io_lock); |
1da177e4 LT |
451 | return 0; /* deferred successfully */ |
452 | } | |
453 | ||
454 | /* | |
455 | * Everyone (including functions in this file), should use this | |
456 | * function to access the md->map field, and make sure they call | |
457 | * dm_table_put() when finished. | |
458 | */ | |
459 | struct dm_table *dm_get_table(struct mapped_device *md) | |
460 | { | |
461 | struct dm_table *t; | |
462 | ||
463 | read_lock(&md->map_lock); | |
464 | t = md->map; | |
465 | if (t) | |
466 | dm_table_get(t); | |
467 | read_unlock(&md->map_lock); | |
468 | ||
469 | return t; | |
470 | } | |
471 | ||
3ac51e74 DW |
472 | /* |
473 | * Get the geometry associated with a dm device | |
474 | */ | |
475 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
476 | { | |
477 | *geo = md->geometry; | |
478 | ||
479 | return 0; | |
480 | } | |
481 | ||
482 | /* | |
483 | * Set the geometry of a device. | |
484 | */ | |
485 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
486 | { | |
487 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
488 | ||
489 | if (geo->start > sz) { | |
490 | DMWARN("Start sector is beyond the geometry limits."); | |
491 | return -EINVAL; | |
492 | } | |
493 | ||
494 | md->geometry = *geo; | |
495 | ||
496 | return 0; | |
497 | } | |
498 | ||
1da177e4 LT |
499 | /*----------------------------------------------------------------- |
500 | * CRUD START: | |
501 | * A more elegant soln is in the works that uses the queue | |
502 | * merge fn, unfortunately there are a couple of changes to | |
503 | * the block layer that I want to make for this. So in the | |
504 | * interests of getting something for people to use I give | |
505 | * you this clearly demarcated crap. | |
506 | *---------------------------------------------------------------*/ | |
507 | ||
2e93ccc1 KU |
508 | static int __noflush_suspending(struct mapped_device *md) |
509 | { | |
510 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
511 | } | |
512 | ||
1da177e4 LT |
513 | /* |
514 | * Decrements the number of outstanding ios that a bio has been | |
515 | * cloned into, completing the original io if necc. | |
516 | */ | |
858119e1 | 517 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 518 | { |
2e93ccc1 | 519 | unsigned long flags; |
b35f8caa MB |
520 | int io_error; |
521 | struct bio *bio; | |
522 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
523 | |
524 | /* Push-back supersedes any I/O errors */ | |
b35f8caa | 525 | if (error && !(io->error > 0 && __noflush_suspending(md))) |
1da177e4 LT |
526 | io->error = error; |
527 | ||
528 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
529 | if (io->error == DM_ENDIO_REQUEUE) { |
530 | /* | |
531 | * Target requested pushing back the I/O. | |
532 | * This must be handled before the sleeper on | |
533 | * suspend queue merges the pushback list. | |
534 | */ | |
b35f8caa MB |
535 | spin_lock_irqsave(&md->pushback_lock, flags); |
536 | if (__noflush_suspending(md)) | |
537 | bio_list_add(&md->pushback, io->bio); | |
2e93ccc1 KU |
538 | else |
539 | /* noflush suspend was interrupted. */ | |
540 | io->error = -EIO; | |
b35f8caa | 541 | spin_unlock_irqrestore(&md->pushback_lock, flags); |
2e93ccc1 KU |
542 | } |
543 | ||
d221d2e7 | 544 | end_io_acct(io); |
1da177e4 | 545 | |
b35f8caa MB |
546 | io_error = io->error; |
547 | bio = io->bio; | |
2e93ccc1 | 548 | |
b35f8caa MB |
549 | free_io(md, io); |
550 | ||
551 | if (io_error != DM_ENDIO_REQUEUE) { | |
552 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 553 | |
b35f8caa MB |
554 | bio_endio(bio, io_error); |
555 | } | |
1da177e4 LT |
556 | } |
557 | } | |
558 | ||
6712ecf8 | 559 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
560 | { |
561 | int r = 0; | |
028867ac | 562 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 563 | struct dm_io *io = tio->io; |
9faf400f | 564 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
565 | dm_endio_fn endio = tio->ti->type->end_io; |
566 | ||
1da177e4 LT |
567 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
568 | error = -EIO; | |
569 | ||
570 | if (endio) { | |
571 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
572 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
573 | /* | |
574 | * error and requeue request are handled | |
575 | * in dec_pending(). | |
576 | */ | |
1da177e4 | 577 | error = r; |
45cbcd79 KU |
578 | else if (r == DM_ENDIO_INCOMPLETE) |
579 | /* The target will handle the io */ | |
6712ecf8 | 580 | return; |
45cbcd79 KU |
581 | else if (r) { |
582 | DMWARN("unimplemented target endio return value: %d", r); | |
583 | BUG(); | |
584 | } | |
1da177e4 LT |
585 | } |
586 | ||
9faf400f SB |
587 | /* |
588 | * Store md for cleanup instead of tio which is about to get freed. | |
589 | */ | |
590 | bio->bi_private = md->bs; | |
591 | ||
9faf400f | 592 | free_tio(md, tio); |
b35f8caa MB |
593 | bio_put(bio); |
594 | dec_pending(io, error); | |
1da177e4 LT |
595 | } |
596 | ||
597 | static sector_t max_io_len(struct mapped_device *md, | |
598 | sector_t sector, struct dm_target *ti) | |
599 | { | |
600 | sector_t offset = sector - ti->begin; | |
601 | sector_t len = ti->len - offset; | |
602 | ||
603 | /* | |
604 | * Does the target need to split even further ? | |
605 | */ | |
606 | if (ti->split_io) { | |
607 | sector_t boundary; | |
608 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
609 | - offset; | |
610 | if (len > boundary) | |
611 | len = boundary; | |
612 | } | |
613 | ||
614 | return len; | |
615 | } | |
616 | ||
617 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 618 | struct dm_target_io *tio) |
1da177e4 LT |
619 | { |
620 | int r; | |
2056a782 | 621 | sector_t sector; |
9faf400f | 622 | struct mapped_device *md; |
1da177e4 LT |
623 | |
624 | /* | |
625 | * Sanity checks. | |
626 | */ | |
627 | BUG_ON(!clone->bi_size); | |
628 | ||
629 | clone->bi_end_io = clone_endio; | |
630 | clone->bi_private = tio; | |
631 | ||
632 | /* | |
633 | * Map the clone. If r == 0 we don't need to do | |
634 | * anything, the target has assumed ownership of | |
635 | * this io. | |
636 | */ | |
637 | atomic_inc(&tio->io->io_count); | |
2056a782 | 638 | sector = clone->bi_sector; |
1da177e4 | 639 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 640 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 641 | /* the bio has been remapped so dispatch it */ |
2056a782 | 642 | |
5f3ea37c | 643 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
644 | tio->io->bio->bi_bdev->bd_dev, |
645 | clone->bi_sector, sector); | |
2056a782 | 646 | |
1da177e4 | 647 | generic_make_request(clone); |
2e93ccc1 KU |
648 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
649 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
650 | md = tio->io->md; |
651 | dec_pending(tio->io, r); | |
652 | /* | |
653 | * Store bio_set for cleanup. | |
654 | */ | |
655 | clone->bi_private = md->bs; | |
1da177e4 | 656 | bio_put(clone); |
9faf400f | 657 | free_tio(md, tio); |
45cbcd79 KU |
658 | } else if (r) { |
659 | DMWARN("unimplemented target map return value: %d", r); | |
660 | BUG(); | |
1da177e4 LT |
661 | } |
662 | } | |
663 | ||
664 | struct clone_info { | |
665 | struct mapped_device *md; | |
666 | struct dm_table *map; | |
667 | struct bio *bio; | |
668 | struct dm_io *io; | |
669 | sector_t sector; | |
670 | sector_t sector_count; | |
671 | unsigned short idx; | |
672 | }; | |
673 | ||
3676347a PO |
674 | static void dm_bio_destructor(struct bio *bio) |
675 | { | |
9faf400f SB |
676 | struct bio_set *bs = bio->bi_private; |
677 | ||
678 | bio_free(bio, bs); | |
3676347a PO |
679 | } |
680 | ||
1da177e4 LT |
681 | /* |
682 | * Creates a little bio that is just does part of a bvec. | |
683 | */ | |
684 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
685 | unsigned short idx, unsigned int offset, | |
9faf400f | 686 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
687 | { |
688 | struct bio *clone; | |
689 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
690 | ||
9faf400f | 691 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 692 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
693 | *clone->bi_io_vec = *bv; |
694 | ||
695 | clone->bi_sector = sector; | |
696 | clone->bi_bdev = bio->bi_bdev; | |
697 | clone->bi_rw = bio->bi_rw; | |
698 | clone->bi_vcnt = 1; | |
699 | clone->bi_size = to_bytes(len); | |
700 | clone->bi_io_vec->bv_offset = offset; | |
701 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 702 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 LT |
703 | |
704 | return clone; | |
705 | } | |
706 | ||
707 | /* | |
708 | * Creates a bio that consists of range of complete bvecs. | |
709 | */ | |
710 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
711 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 712 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
713 | { |
714 | struct bio *clone; | |
715 | ||
9faf400f SB |
716 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
717 | __bio_clone(clone, bio); | |
718 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
719 | clone->bi_sector = sector; |
720 | clone->bi_idx = idx; | |
721 | clone->bi_vcnt = idx + bv_count; | |
722 | clone->bi_size = to_bytes(len); | |
723 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
724 | ||
725 | return clone; | |
726 | } | |
727 | ||
512875bd | 728 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
729 | { |
730 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
731 | struct dm_target *ti; |
732 | sector_t len = 0, max; | |
028867ac | 733 | struct dm_target_io *tio; |
1da177e4 | 734 | |
512875bd JN |
735 | ti = dm_table_find_target(ci->map, ci->sector); |
736 | if (!dm_target_is_valid(ti)) | |
737 | return -EIO; | |
738 | ||
739 | max = max_io_len(ci->md, ci->sector, ti); | |
740 | ||
1da177e4 LT |
741 | /* |
742 | * Allocate a target io object. | |
743 | */ | |
744 | tio = alloc_tio(ci->md); | |
745 | tio->io = ci->io; | |
746 | tio->ti = ti; | |
747 | memset(&tio->info, 0, sizeof(tio->info)); | |
748 | ||
749 | if (ci->sector_count <= max) { | |
750 | /* | |
751 | * Optimise for the simple case where we can do all of | |
752 | * the remaining io with a single clone. | |
753 | */ | |
754 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
755 | bio->bi_vcnt - ci->idx, ci->sector_count, |
756 | ci->md->bs); | |
1da177e4 LT |
757 | __map_bio(ti, clone, tio); |
758 | ci->sector_count = 0; | |
759 | ||
760 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
761 | /* | |
762 | * There are some bvecs that don't span targets. | |
763 | * Do as many of these as possible. | |
764 | */ | |
765 | int i; | |
766 | sector_t remaining = max; | |
767 | sector_t bv_len; | |
768 | ||
769 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
770 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
771 | ||
772 | if (bv_len > remaining) | |
773 | break; | |
774 | ||
775 | remaining -= bv_len; | |
776 | len += bv_len; | |
777 | } | |
778 | ||
9faf400f SB |
779 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
780 | ci->md->bs); | |
1da177e4 LT |
781 | __map_bio(ti, clone, tio); |
782 | ||
783 | ci->sector += len; | |
784 | ci->sector_count -= len; | |
785 | ci->idx = i; | |
786 | ||
787 | } else { | |
788 | /* | |
d2044a94 | 789 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
790 | */ |
791 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
792 | sector_t remaining = to_sector(bv->bv_len); |
793 | unsigned int offset = 0; | |
1da177e4 | 794 | |
d2044a94 AK |
795 | do { |
796 | if (offset) { | |
797 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
798 | if (!dm_target_is_valid(ti)) |
799 | return -EIO; | |
800 | ||
d2044a94 | 801 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 802 | |
d2044a94 AK |
803 | tio = alloc_tio(ci->md); |
804 | tio->io = ci->io; | |
805 | tio->ti = ti; | |
806 | memset(&tio->info, 0, sizeof(tio->info)); | |
807 | } | |
808 | ||
809 | len = min(remaining, max); | |
810 | ||
811 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
812 | bv->bv_offset + offset, len, |
813 | ci->md->bs); | |
d2044a94 AK |
814 | |
815 | __map_bio(ti, clone, tio); | |
816 | ||
817 | ci->sector += len; | |
818 | ci->sector_count -= len; | |
819 | offset += to_bytes(len); | |
820 | } while (remaining -= len); | |
1da177e4 | 821 | |
1da177e4 LT |
822 | ci->idx++; |
823 | } | |
512875bd JN |
824 | |
825 | return 0; | |
1da177e4 LT |
826 | } |
827 | ||
828 | /* | |
8a53c28d | 829 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 830 | */ |
f0b9a450 | 831 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
832 | { |
833 | struct clone_info ci; | |
512875bd | 834 | int error = 0; |
1da177e4 LT |
835 | |
836 | ci.map = dm_get_table(md); | |
f0b9a450 MP |
837 | if (unlikely(!ci.map)) { |
838 | bio_io_error(bio); | |
839 | return; | |
840 | } | |
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); | |
f0b9a450 | 844 | return; |
ab4c1424 | 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 LT |
863 | dm_table_put(ci.map); |
864 | } | |
865 | /*----------------------------------------------------------------- | |
866 | * CRUD END | |
867 | *---------------------------------------------------------------*/ | |
868 | ||
f6fccb12 MB |
869 | static int dm_merge_bvec(struct request_queue *q, |
870 | struct bvec_merge_data *bvm, | |
871 | struct bio_vec *biovec) | |
872 | { | |
873 | struct mapped_device *md = q->queuedata; | |
874 | struct dm_table *map = dm_get_table(md); | |
875 | struct dm_target *ti; | |
876 | sector_t max_sectors; | |
5037108a | 877 | int max_size = 0; |
f6fccb12 MB |
878 | |
879 | if (unlikely(!map)) | |
5037108a | 880 | goto out; |
f6fccb12 MB |
881 | |
882 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
883 | if (!dm_target_is_valid(ti)) |
884 | goto out_table; | |
f6fccb12 MB |
885 | |
886 | /* | |
887 | * Find maximum amount of I/O that won't need splitting | |
888 | */ | |
889 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
890 | (sector_t) BIO_MAX_SECTORS); | |
891 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
892 | if (max_size < 0) | |
893 | max_size = 0; | |
894 | ||
895 | /* | |
896 | * merge_bvec_fn() returns number of bytes | |
897 | * it can accept at this offset | |
898 | * max is precomputed maximal io size | |
899 | */ | |
900 | if (max_size && ti->type->merge) | |
901 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
902 | ||
b01cd5ac | 903 | out_table: |
5037108a MP |
904 | dm_table_put(map); |
905 | ||
906 | out: | |
f6fccb12 MB |
907 | /* |
908 | * Always allow an entire first page | |
909 | */ | |
910 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
911 | max_size = biovec->bv_len; | |
912 | ||
f6fccb12 MB |
913 | return max_size; |
914 | } | |
915 | ||
1da177e4 LT |
916 | /* |
917 | * The request function that just remaps the bio built up by | |
918 | * dm_merge_bvec. | |
919 | */ | |
165125e1 | 920 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 921 | { |
9e4e5f87 | 922 | int r = -EIO; |
12f03a49 | 923 | int rw = bio_data_dir(bio); |
1da177e4 | 924 | struct mapped_device *md = q->queuedata; |
c9959059 | 925 | int cpu; |
1da177e4 | 926 | |
2ca3310e | 927 | down_read(&md->io_lock); |
1da177e4 | 928 | |
074a7aca TH |
929 | cpu = part_stat_lock(); |
930 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
931 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
932 | part_stat_unlock(); | |
12f03a49 | 933 | |
1da177e4 LT |
934 | /* |
935 | * If we're suspended we have to queue | |
936 | * this io for later. | |
937 | */ | |
938 | while (test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 939 | up_read(&md->io_lock); |
1da177e4 | 940 | |
9e4e5f87 MB |
941 | if (bio_rw(bio) != READA) |
942 | r = queue_io(md, bio); | |
1da177e4 | 943 | |
9e4e5f87 MB |
944 | if (r <= 0) |
945 | goto out_req; | |
1da177e4 LT |
946 | |
947 | /* | |
948 | * We're in a while loop, because someone could suspend | |
949 | * before we get to the following read lock. | |
950 | */ | |
2ca3310e | 951 | down_read(&md->io_lock); |
1da177e4 LT |
952 | } |
953 | ||
f0b9a450 | 954 | __split_and_process_bio(md, bio); |
2ca3310e | 955 | up_read(&md->io_lock); |
f0b9a450 | 956 | return 0; |
9e4e5f87 MB |
957 | |
958 | out_req: | |
959 | if (r < 0) | |
960 | bio_io_error(bio); | |
961 | ||
1da177e4 LT |
962 | return 0; |
963 | } | |
964 | ||
165125e1 | 965 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
966 | { |
967 | struct mapped_device *md = q->queuedata; | |
968 | struct dm_table *map = dm_get_table(md); | |
969 | ||
970 | if (map) { | |
971 | dm_table_unplug_all(map); | |
972 | dm_table_put(map); | |
973 | } | |
974 | } | |
975 | ||
976 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
977 | { | |
8a57dfc6 CS |
978 | int r = bdi_bits; |
979 | struct mapped_device *md = congested_data; | |
980 | struct dm_table *map; | |
1da177e4 | 981 | |
8a57dfc6 CS |
982 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { |
983 | map = dm_get_table(md); | |
984 | if (map) { | |
985 | r = dm_table_any_congested(map, bdi_bits); | |
986 | dm_table_put(map); | |
987 | } | |
988 | } | |
989 | ||
1da177e4 LT |
990 | return r; |
991 | } | |
992 | ||
993 | /*----------------------------------------------------------------- | |
994 | * An IDR is used to keep track of allocated minor numbers. | |
995 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
996 | static DEFINE_IDR(_minor_idr); |
997 | ||
2b06cfff | 998 | static void free_minor(int minor) |
1da177e4 | 999 | { |
f32c10b0 | 1000 | spin_lock(&_minor_lock); |
1da177e4 | 1001 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1002 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1003 | } |
1004 | ||
1005 | /* | |
1006 | * See if the device with a specific minor # is free. | |
1007 | */ | |
cf13ab8e | 1008 | static int specific_minor(int minor) |
1da177e4 LT |
1009 | { |
1010 | int r, m; | |
1011 | ||
1012 | if (minor >= (1 << MINORBITS)) | |
1013 | return -EINVAL; | |
1014 | ||
62f75c2f JM |
1015 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1016 | if (!r) | |
1017 | return -ENOMEM; | |
1018 | ||
f32c10b0 | 1019 | spin_lock(&_minor_lock); |
1da177e4 LT |
1020 | |
1021 | if (idr_find(&_minor_idr, minor)) { | |
1022 | r = -EBUSY; | |
1023 | goto out; | |
1024 | } | |
1025 | ||
ba61fdd1 | 1026 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1027 | if (r) |
1da177e4 | 1028 | goto out; |
1da177e4 LT |
1029 | |
1030 | if (m != minor) { | |
1031 | idr_remove(&_minor_idr, m); | |
1032 | r = -EBUSY; | |
1033 | goto out; | |
1034 | } | |
1035 | ||
1036 | out: | |
f32c10b0 | 1037 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1038 | return r; |
1039 | } | |
1040 | ||
cf13ab8e | 1041 | static int next_free_minor(int *minor) |
1da177e4 | 1042 | { |
2b06cfff | 1043 | int r, m; |
1da177e4 | 1044 | |
1da177e4 | 1045 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1046 | if (!r) |
1047 | return -ENOMEM; | |
1048 | ||
f32c10b0 | 1049 | spin_lock(&_minor_lock); |
1da177e4 | 1050 | |
ba61fdd1 | 1051 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1052 | if (r) |
1da177e4 | 1053 | goto out; |
1da177e4 LT |
1054 | |
1055 | if (m >= (1 << MINORBITS)) { | |
1056 | idr_remove(&_minor_idr, m); | |
1057 | r = -ENOSPC; | |
1058 | goto out; | |
1059 | } | |
1060 | ||
1061 | *minor = m; | |
1062 | ||
1063 | out: | |
f32c10b0 | 1064 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1065 | return r; |
1066 | } | |
1067 | ||
1068 | static struct block_device_operations dm_blk_dops; | |
1069 | ||
53d5914f MP |
1070 | static void dm_wq_work(struct work_struct *work); |
1071 | ||
1da177e4 LT |
1072 | /* |
1073 | * Allocate and initialise a blank device with a given minor. | |
1074 | */ | |
2b06cfff | 1075 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1076 | { |
1077 | int r; | |
cf13ab8e | 1078 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1079 | void *old_md; |
1da177e4 LT |
1080 | |
1081 | if (!md) { | |
1082 | DMWARN("unable to allocate device, out of memory."); | |
1083 | return NULL; | |
1084 | } | |
1085 | ||
10da4f79 | 1086 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1087 | goto bad_module_get; |
10da4f79 | 1088 | |
1da177e4 | 1089 | /* get a minor number for the dev */ |
2b06cfff | 1090 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1091 | r = next_free_minor(&minor); |
2b06cfff | 1092 | else |
cf13ab8e | 1093 | r = specific_minor(minor); |
1da177e4 | 1094 | if (r < 0) |
6ed7ade8 | 1095 | goto bad_minor; |
1da177e4 | 1096 | |
2ca3310e | 1097 | init_rwsem(&md->io_lock); |
e61290a4 | 1098 | mutex_init(&md->suspend_lock); |
2e93ccc1 | 1099 | spin_lock_init(&md->pushback_lock); |
1da177e4 LT |
1100 | rwlock_init(&md->map_lock); |
1101 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1102 | atomic_set(&md->open_count, 0); |
1da177e4 | 1103 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1104 | atomic_set(&md->uevent_seq, 0); |
1105 | INIT_LIST_HEAD(&md->uevent_list); | |
1106 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1107 | |
1108 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1109 | if (!md->queue) | |
6ed7ade8 | 1110 | goto bad_queue; |
1da177e4 LT |
1111 | |
1112 | md->queue->queuedata = md; | |
1113 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1114 | md->queue->backing_dev_info.congested_data = md; | |
1115 | blk_queue_make_request(md->queue, dm_request); | |
daef265f | 1116 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1117 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1118 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1119 | |
93d2341c | 1120 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1121 | if (!md->io_pool) |
6ed7ade8 | 1122 | goto bad_io_pool; |
1da177e4 | 1123 | |
93d2341c | 1124 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1125 | if (!md->tio_pool) |
6ed7ade8 | 1126 | goto bad_tio_pool; |
1da177e4 | 1127 | |
bb799ca0 | 1128 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1129 | if (!md->bs) |
1130 | goto bad_no_bioset; | |
1131 | ||
1da177e4 LT |
1132 | md->disk = alloc_disk(1); |
1133 | if (!md->disk) | |
6ed7ade8 | 1134 | goto bad_disk; |
1da177e4 | 1135 | |
f0b04115 JM |
1136 | atomic_set(&md->pending, 0); |
1137 | init_waitqueue_head(&md->wait); | |
53d5914f | 1138 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1139 | init_waitqueue_head(&md->eventq); |
1140 | ||
1da177e4 LT |
1141 | md->disk->major = _major; |
1142 | md->disk->first_minor = minor; | |
1143 | md->disk->fops = &dm_blk_dops; | |
1144 | md->disk->queue = md->queue; | |
1145 | md->disk->private_data = md; | |
1146 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1147 | add_disk(md->disk); | |
7e51f257 | 1148 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1149 | |
304f3f6a MB |
1150 | md->wq = create_singlethread_workqueue("kdmflush"); |
1151 | if (!md->wq) | |
1152 | goto bad_thread; | |
1153 | ||
ba61fdd1 | 1154 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1155 | spin_lock(&_minor_lock); |
ba61fdd1 | 1156 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1157 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1158 | |
1159 | BUG_ON(old_md != MINOR_ALLOCED); | |
1160 | ||
1da177e4 LT |
1161 | return md; |
1162 | ||
304f3f6a MB |
1163 | bad_thread: |
1164 | put_disk(md->disk); | |
6ed7ade8 | 1165 | bad_disk: |
9faf400f | 1166 | bioset_free(md->bs); |
6ed7ade8 | 1167 | bad_no_bioset: |
1da177e4 | 1168 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1169 | bad_tio_pool: |
1da177e4 | 1170 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1171 | bad_io_pool: |
1312f40e | 1172 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1173 | bad_queue: |
1da177e4 | 1174 | free_minor(minor); |
6ed7ade8 | 1175 | bad_minor: |
10da4f79 | 1176 | module_put(THIS_MODULE); |
6ed7ade8 | 1177 | bad_module_get: |
1da177e4 LT |
1178 | kfree(md); |
1179 | return NULL; | |
1180 | } | |
1181 | ||
ae9da83f JN |
1182 | static void unlock_fs(struct mapped_device *md); |
1183 | ||
1da177e4 LT |
1184 | static void free_dev(struct mapped_device *md) |
1185 | { | |
f331c029 | 1186 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1187 | |
d9dde59b | 1188 | if (md->suspended_bdev) { |
ae9da83f | 1189 | unlock_fs(md); |
d9dde59b JN |
1190 | bdput(md->suspended_bdev); |
1191 | } | |
304f3f6a | 1192 | destroy_workqueue(md->wq); |
1da177e4 LT |
1193 | mempool_destroy(md->tio_pool); |
1194 | mempool_destroy(md->io_pool); | |
9faf400f | 1195 | bioset_free(md->bs); |
1da177e4 | 1196 | del_gendisk(md->disk); |
63d94e48 | 1197 | free_minor(minor); |
fba9f90e JM |
1198 | |
1199 | spin_lock(&_minor_lock); | |
1200 | md->disk->private_data = NULL; | |
1201 | spin_unlock(&_minor_lock); | |
1202 | ||
1da177e4 | 1203 | put_disk(md->disk); |
1312f40e | 1204 | blk_cleanup_queue(md->queue); |
10da4f79 | 1205 | module_put(THIS_MODULE); |
1da177e4 LT |
1206 | kfree(md); |
1207 | } | |
1208 | ||
1209 | /* | |
1210 | * Bind a table to the device. | |
1211 | */ | |
1212 | static void event_callback(void *context) | |
1213 | { | |
7a8c3d3b MA |
1214 | unsigned long flags; |
1215 | LIST_HEAD(uevents); | |
1da177e4 LT |
1216 | struct mapped_device *md = (struct mapped_device *) context; |
1217 | ||
7a8c3d3b MA |
1218 | spin_lock_irqsave(&md->uevent_lock, flags); |
1219 | list_splice_init(&md->uevent_list, &uevents); | |
1220 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1221 | ||
ed9e1982 | 1222 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1223 | |
1da177e4 LT |
1224 | atomic_inc(&md->event_nr); |
1225 | wake_up(&md->eventq); | |
1226 | } | |
1227 | ||
4e90188b | 1228 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1229 | { |
4e90188b | 1230 | set_capacity(md->disk, size); |
1da177e4 | 1231 | |
1b1dcc1b | 1232 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1233 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1234 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1235 | } |
1236 | ||
1237 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1238 | { | |
165125e1 | 1239 | struct request_queue *q = md->queue; |
1da177e4 LT |
1240 | sector_t size; |
1241 | ||
1242 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1243 | |
1244 | /* | |
1245 | * Wipe any geometry if the size of the table changed. | |
1246 | */ | |
1247 | if (size != get_capacity(md->disk)) | |
1248 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1249 | ||
bfa152fa JN |
1250 | if (md->suspended_bdev) |
1251 | __set_size(md, size); | |
d5816876 MP |
1252 | |
1253 | if (!size) { | |
1254 | dm_table_destroy(t); | |
1da177e4 | 1255 | return 0; |
d5816876 | 1256 | } |
1da177e4 | 1257 | |
2ca3310e AK |
1258 | dm_table_event_callback(t, event_callback, md); |
1259 | ||
1da177e4 LT |
1260 | write_lock(&md->map_lock); |
1261 | md->map = t; | |
2ca3310e | 1262 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1263 | write_unlock(&md->map_lock); |
1264 | ||
1da177e4 LT |
1265 | return 0; |
1266 | } | |
1267 | ||
1268 | static void __unbind(struct mapped_device *md) | |
1269 | { | |
1270 | struct dm_table *map = md->map; | |
1271 | ||
1272 | if (!map) | |
1273 | return; | |
1274 | ||
1275 | dm_table_event_callback(map, NULL, NULL); | |
1276 | write_lock(&md->map_lock); | |
1277 | md->map = NULL; | |
1278 | write_unlock(&md->map_lock); | |
d5816876 | 1279 | dm_table_destroy(map); |
1da177e4 LT |
1280 | } |
1281 | ||
1282 | /* | |
1283 | * Constructor for a new device. | |
1284 | */ | |
2b06cfff | 1285 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1286 | { |
1287 | struct mapped_device *md; | |
1288 | ||
2b06cfff | 1289 | md = alloc_dev(minor); |
1da177e4 LT |
1290 | if (!md) |
1291 | return -ENXIO; | |
1292 | ||
784aae73 MB |
1293 | dm_sysfs_init(md); |
1294 | ||
1da177e4 LT |
1295 | *result = md; |
1296 | return 0; | |
1297 | } | |
1298 | ||
637842cf | 1299 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1300 | { |
1301 | struct mapped_device *md; | |
1da177e4 LT |
1302 | unsigned minor = MINOR(dev); |
1303 | ||
1304 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1305 | return NULL; | |
1306 | ||
f32c10b0 | 1307 | spin_lock(&_minor_lock); |
1da177e4 LT |
1308 | |
1309 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1310 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1311 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1312 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1313 | md = NULL; |
fba9f90e JM |
1314 | goto out; |
1315 | } | |
1da177e4 | 1316 | |
fba9f90e | 1317 | out: |
f32c10b0 | 1318 | spin_unlock(&_minor_lock); |
1da177e4 | 1319 | |
637842cf DT |
1320 | return md; |
1321 | } | |
1322 | ||
d229a958 DT |
1323 | struct mapped_device *dm_get_md(dev_t dev) |
1324 | { | |
1325 | struct mapped_device *md = dm_find_md(dev); | |
1326 | ||
1327 | if (md) | |
1328 | dm_get(md); | |
1329 | ||
1330 | return md; | |
1331 | } | |
1332 | ||
9ade92a9 | 1333 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1334 | { |
9ade92a9 | 1335 | return md->interface_ptr; |
1da177e4 LT |
1336 | } |
1337 | ||
1338 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1339 | { | |
1340 | md->interface_ptr = ptr; | |
1341 | } | |
1342 | ||
1343 | void dm_get(struct mapped_device *md) | |
1344 | { | |
1345 | atomic_inc(&md->holders); | |
1346 | } | |
1347 | ||
72d94861 AK |
1348 | const char *dm_device_name(struct mapped_device *md) |
1349 | { | |
1350 | return md->name; | |
1351 | } | |
1352 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1353 | ||
1da177e4 LT |
1354 | void dm_put(struct mapped_device *md) |
1355 | { | |
1134e5ae | 1356 | struct dm_table *map; |
1da177e4 | 1357 | |
fba9f90e JM |
1358 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1359 | ||
f32c10b0 | 1360 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1361 | map = dm_get_table(md); |
f331c029 TH |
1362 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1363 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1364 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1365 | spin_unlock(&_minor_lock); |
cf222b37 | 1366 | if (!dm_suspended(md)) { |
1da177e4 LT |
1367 | dm_table_presuspend_targets(map); |
1368 | dm_table_postsuspend_targets(map); | |
1369 | } | |
784aae73 | 1370 | dm_sysfs_exit(md); |
1134e5ae | 1371 | dm_table_put(map); |
a1b51e98 | 1372 | __unbind(md); |
1da177e4 LT |
1373 | free_dev(md); |
1374 | } | |
1da177e4 | 1375 | } |
79eb885c | 1376 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1377 | |
46125c1c MB |
1378 | static int dm_wait_for_completion(struct mapped_device *md) |
1379 | { | |
1380 | int r = 0; | |
1381 | ||
1382 | while (1) { | |
1383 | set_current_state(TASK_INTERRUPTIBLE); | |
1384 | ||
1385 | smp_mb(); | |
1386 | if (!atomic_read(&md->pending)) | |
1387 | break; | |
1388 | ||
1389 | if (signal_pending(current)) { | |
1390 | r = -EINTR; | |
1391 | break; | |
1392 | } | |
1393 | ||
1394 | io_schedule(); | |
1395 | } | |
1396 | set_current_state(TASK_RUNNING); | |
1397 | ||
1398 | return r; | |
1399 | } | |
1400 | ||
1da177e4 LT |
1401 | /* |
1402 | * Process the deferred bios | |
1403 | */ | |
ef208587 | 1404 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 1405 | { |
ef208587 MP |
1406 | struct mapped_device *md = container_of(work, struct mapped_device, |
1407 | work); | |
6d6f10df | 1408 | struct bio *c; |
1da177e4 | 1409 | |
ef208587 MP |
1410 | down_write(&md->io_lock); |
1411 | ||
f0b9a450 MP |
1412 | while ((c = bio_list_pop(&md->deferred))) |
1413 | __split_and_process_bio(md, c); | |
73d410c0 MB |
1414 | |
1415 | clear_bit(DMF_BLOCK_IO, &md->flags); | |
ef208587 MP |
1416 | |
1417 | up_write(&md->io_lock); | |
1da177e4 LT |
1418 | } |
1419 | ||
6d6f10df MB |
1420 | static void __merge_pushback_list(struct mapped_device *md) |
1421 | { | |
1422 | unsigned long flags; | |
1423 | ||
1424 | spin_lock_irqsave(&md->pushback_lock, flags); | |
1425 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1426 | bio_list_merge_head(&md->deferred, &md->pushback); | |
1427 | bio_list_init(&md->pushback); | |
1428 | spin_unlock_irqrestore(&md->pushback_lock, flags); | |
1429 | } | |
1430 | ||
9a1fb464 | 1431 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 1432 | { |
53d5914f | 1433 | queue_work(md->wq, &md->work); |
304f3f6a MB |
1434 | flush_workqueue(md->wq); |
1435 | } | |
1436 | ||
1da177e4 LT |
1437 | /* |
1438 | * Swap in a new table (destroying old one). | |
1439 | */ | |
1440 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1441 | { | |
93c534ae | 1442 | int r = -EINVAL; |
1da177e4 | 1443 | |
e61290a4 | 1444 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1445 | |
1446 | /* device must be suspended */ | |
cf222b37 | 1447 | if (!dm_suspended(md)) |
93c534ae | 1448 | goto out; |
1da177e4 | 1449 | |
bfa152fa JN |
1450 | /* without bdev, the device size cannot be changed */ |
1451 | if (!md->suspended_bdev) | |
1452 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1453 | goto out; | |
1454 | ||
1da177e4 LT |
1455 | __unbind(md); |
1456 | r = __bind(md, table); | |
1da177e4 | 1457 | |
93c534ae | 1458 | out: |
e61290a4 | 1459 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1460 | return r; |
1da177e4 LT |
1461 | } |
1462 | ||
1463 | /* | |
1464 | * Functions to lock and unlock any filesystem running on the | |
1465 | * device. | |
1466 | */ | |
2ca3310e | 1467 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1468 | { |
e39e2e95 | 1469 | int r; |
1da177e4 LT |
1470 | |
1471 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1472 | |
e39e2e95 | 1473 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1474 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1475 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1476 | md->frozen_sb = NULL; |
1477 | return r; | |
dfbe03f6 AK |
1478 | } |
1479 | ||
aa8d7c2f AK |
1480 | set_bit(DMF_FROZEN, &md->flags); |
1481 | ||
1da177e4 | 1482 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1483 | * to go away while it is locked. |
1da177e4 LT |
1484 | */ |
1485 | return 0; | |
1486 | } | |
1487 | ||
2ca3310e | 1488 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1489 | { |
aa8d7c2f AK |
1490 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1491 | return; | |
1492 | ||
e39e2e95 | 1493 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1494 | md->frozen_sb = NULL; |
aa8d7c2f | 1495 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1496 | } |
1497 | ||
1498 | /* | |
1499 | * We need to be able to change a mapping table under a mounted | |
1500 | * filesystem. For example we might want to move some data in | |
1501 | * the background. Before the table can be swapped with | |
1502 | * dm_bind_table, dm_suspend must be called to flush any in | |
1503 | * flight bios and ensure that any further io gets deferred. | |
1504 | */ | |
a3d77d35 | 1505 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1506 | { |
2ca3310e | 1507 | struct dm_table *map = NULL; |
1da177e4 | 1508 | DECLARE_WAITQUEUE(wait, current); |
46125c1c | 1509 | int r = 0; |
a3d77d35 | 1510 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1511 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1512 | |
e61290a4 | 1513 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1514 | |
73d410c0 MB |
1515 | if (dm_suspended(md)) { |
1516 | r = -EINVAL; | |
d287483d | 1517 | goto out_unlock; |
73d410c0 | 1518 | } |
1da177e4 LT |
1519 | |
1520 | map = dm_get_table(md); | |
1da177e4 | 1521 | |
2e93ccc1 KU |
1522 | /* |
1523 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1524 | * This flag is cleared before dm_suspend returns. | |
1525 | */ | |
1526 | if (noflush) | |
1527 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1528 | ||
cf222b37 AK |
1529 | /* This does not get reverted if there's an error later. */ |
1530 | dm_table_presuspend_targets(map); | |
1531 | ||
bfa152fa JN |
1532 | /* bdget() can stall if the pending I/Os are not flushed */ |
1533 | if (!noflush) { | |
1534 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1535 | if (!md->suspended_bdev) { | |
1536 | DMWARN("bdget failed in dm_suspend"); | |
1537 | r = -ENOMEM; | |
f431d966 | 1538 | goto out; |
bfa152fa | 1539 | } |
e39e2e95 | 1540 | |
6d6f10df MB |
1541 | /* |
1542 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1543 | * because lock_fs() needs to flush I/Os. | |
1544 | */ | |
1545 | if (do_lockfs) { | |
1546 | r = lock_fs(md); | |
1547 | if (r) | |
1548 | goto out; | |
1549 | } | |
aa8d7c2f | 1550 | } |
1da177e4 LT |
1551 | |
1552 | /* | |
354e0071 | 1553 | * First we set the BLOCK_IO flag so no more ios will be mapped. |
1da177e4 | 1554 | */ |
2ca3310e AK |
1555 | down_write(&md->io_lock); |
1556 | set_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 | 1557 | |
1da177e4 | 1558 | add_wait_queue(&md->wait, &wait); |
2ca3310e | 1559 | up_write(&md->io_lock); |
1da177e4 LT |
1560 | |
1561 | /* unplug */ | |
2ca3310e | 1562 | if (map) |
1da177e4 | 1563 | dm_table_unplug_all(map); |
1da177e4 LT |
1564 | |
1565 | /* | |
46125c1c | 1566 | * Wait for the already-mapped ios to complete. |
1da177e4 | 1567 | */ |
46125c1c | 1568 | r = dm_wait_for_completion(md); |
1da177e4 | 1569 | |
2ca3310e | 1570 | down_write(&md->io_lock); |
1da177e4 LT |
1571 | remove_wait_queue(&md->wait, &wait); |
1572 | ||
6d6f10df MB |
1573 | if (noflush) |
1574 | __merge_pushback_list(md); | |
94d6351e | 1575 | up_write(&md->io_lock); |
2e93ccc1 | 1576 | |
1da177e4 | 1577 | /* were we interrupted ? */ |
46125c1c | 1578 | if (r < 0) { |
9a1fb464 | 1579 | dm_queue_flush(md); |
73d410c0 | 1580 | |
2ca3310e | 1581 | unlock_fs(md); |
2e93ccc1 | 1582 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1583 | } |
1da177e4 | 1584 | |
cf222b37 | 1585 | dm_table_postsuspend_targets(map); |
1da177e4 | 1586 | |
2ca3310e | 1587 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1588 | |
2ca3310e | 1589 | out: |
e39e2e95 AK |
1590 | if (r && md->suspended_bdev) { |
1591 | bdput(md->suspended_bdev); | |
1592 | md->suspended_bdev = NULL; | |
1593 | } | |
1594 | ||
2ca3310e | 1595 | dm_table_put(map); |
d287483d AK |
1596 | |
1597 | out_unlock: | |
e61290a4 | 1598 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1599 | return r; |
1da177e4 LT |
1600 | } |
1601 | ||
1602 | int dm_resume(struct mapped_device *md) | |
1603 | { | |
cf222b37 | 1604 | int r = -EINVAL; |
cf222b37 | 1605 | struct dm_table *map = NULL; |
1da177e4 | 1606 | |
e61290a4 | 1607 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1608 | if (!dm_suspended(md)) |
cf222b37 | 1609 | goto out; |
cf222b37 AK |
1610 | |
1611 | map = dm_get_table(md); | |
2ca3310e | 1612 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1613 | goto out; |
1da177e4 | 1614 | |
8757b776 MB |
1615 | r = dm_table_resume_targets(map); |
1616 | if (r) | |
1617 | goto out; | |
2ca3310e | 1618 | |
9a1fb464 | 1619 | dm_queue_flush(md); |
2ca3310e AK |
1620 | |
1621 | unlock_fs(md); | |
1622 | ||
bfa152fa JN |
1623 | if (md->suspended_bdev) { |
1624 | bdput(md->suspended_bdev); | |
1625 | md->suspended_bdev = NULL; | |
1626 | } | |
e39e2e95 | 1627 | |
2ca3310e AK |
1628 | clear_bit(DMF_SUSPENDED, &md->flags); |
1629 | ||
1da177e4 | 1630 | dm_table_unplug_all(map); |
1da177e4 | 1631 | |
69267a30 | 1632 | dm_kobject_uevent(md); |
8560ed6f | 1633 | |
cf222b37 | 1634 | r = 0; |
2ca3310e | 1635 | |
cf222b37 AK |
1636 | out: |
1637 | dm_table_put(map); | |
e61290a4 | 1638 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1639 | |
cf222b37 | 1640 | return r; |
1da177e4 LT |
1641 | } |
1642 | ||
1643 | /*----------------------------------------------------------------- | |
1644 | * Event notification. | |
1645 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1646 | void dm_kobject_uevent(struct mapped_device *md) |
1647 | { | |
ed9e1982 | 1648 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1649 | } |
1650 | ||
7a8c3d3b MA |
1651 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1652 | { | |
1653 | return atomic_add_return(1, &md->uevent_seq); | |
1654 | } | |
1655 | ||
1da177e4 LT |
1656 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1657 | { | |
1658 | return atomic_read(&md->event_nr); | |
1659 | } | |
1660 | ||
1661 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1662 | { | |
1663 | return wait_event_interruptible(md->eventq, | |
1664 | (event_nr != atomic_read(&md->event_nr))); | |
1665 | } | |
1666 | ||
7a8c3d3b MA |
1667 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1668 | { | |
1669 | unsigned long flags; | |
1670 | ||
1671 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1672 | list_add(elist, &md->uevent_list); | |
1673 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1674 | } | |
1675 | ||
1da177e4 LT |
1676 | /* |
1677 | * The gendisk is only valid as long as you have a reference | |
1678 | * count on 'md'. | |
1679 | */ | |
1680 | struct gendisk *dm_disk(struct mapped_device *md) | |
1681 | { | |
1682 | return md->disk; | |
1683 | } | |
1684 | ||
784aae73 MB |
1685 | struct kobject *dm_kobject(struct mapped_device *md) |
1686 | { | |
1687 | return &md->kobj; | |
1688 | } | |
1689 | ||
1690 | /* | |
1691 | * struct mapped_device should not be exported outside of dm.c | |
1692 | * so use this check to verify that kobj is part of md structure | |
1693 | */ | |
1694 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1695 | { | |
1696 | struct mapped_device *md; | |
1697 | ||
1698 | md = container_of(kobj, struct mapped_device, kobj); | |
1699 | if (&md->kobj != kobj) | |
1700 | return NULL; | |
1701 | ||
1702 | dm_get(md); | |
1703 | return md; | |
1704 | } | |
1705 | ||
1da177e4 LT |
1706 | int dm_suspended(struct mapped_device *md) |
1707 | { | |
1708 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1709 | } | |
1710 | ||
2e93ccc1 KU |
1711 | int dm_noflush_suspending(struct dm_target *ti) |
1712 | { | |
1713 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1714 | int r = __noflush_suspending(md); | |
1715 | ||
1716 | dm_put(md); | |
1717 | ||
1718 | return r; | |
1719 | } | |
1720 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1721 | ||
1da177e4 LT |
1722 | static struct block_device_operations dm_blk_dops = { |
1723 | .open = dm_blk_open, | |
1724 | .release = dm_blk_close, | |
aa129a22 | 1725 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1726 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1727 | .owner = THIS_MODULE |
1728 | }; | |
1729 | ||
1730 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1731 | ||
1732 | /* | |
1733 | * module hooks | |
1734 | */ | |
1735 | module_init(dm_init); | |
1736 | module_exit(dm_exit); | |
1737 | ||
1738 | module_param(major, uint, 0); | |
1739 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1740 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1741 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1742 | MODULE_LICENSE("GPL"); |