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; |
1da177e4 LT |
105 | rwlock_t map_lock; |
106 | atomic_t holders; | |
5c6bd75d | 107 | atomic_t open_count; |
1da177e4 LT |
108 | |
109 | unsigned long flags; | |
110 | ||
165125e1 | 111 | struct request_queue *queue; |
1da177e4 | 112 | struct gendisk *disk; |
7e51f257 | 113 | char name[16]; |
1da177e4 LT |
114 | |
115 | void *interface_ptr; | |
116 | ||
117 | /* | |
118 | * A list of ios that arrived while we were suspended. | |
119 | */ | |
120 | atomic_t pending; | |
121 | wait_queue_head_t wait; | |
53d5914f | 122 | struct work_struct work; |
74859364 | 123 | struct bio_list deferred; |
022c2611 | 124 | spinlock_t deferred_lock; |
1da177e4 | 125 | |
304f3f6a MB |
126 | /* |
127 | * Processing queue (flush/barriers) | |
128 | */ | |
129 | struct workqueue_struct *wq; | |
130 | ||
1da177e4 LT |
131 | /* |
132 | * The current mapping. | |
133 | */ | |
134 | struct dm_table *map; | |
135 | ||
136 | /* | |
137 | * io objects are allocated from here. | |
138 | */ | |
139 | mempool_t *io_pool; | |
140 | mempool_t *tio_pool; | |
141 | ||
9faf400f SB |
142 | struct bio_set *bs; |
143 | ||
1da177e4 LT |
144 | /* |
145 | * Event handling. | |
146 | */ | |
147 | atomic_t event_nr; | |
148 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
149 | atomic_t uevent_seq; |
150 | struct list_head uevent_list; | |
151 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
152 | |
153 | /* | |
154 | * freeze/thaw support require holding onto a super block | |
155 | */ | |
156 | struct super_block *frozen_sb; | |
e39e2e95 | 157 | struct block_device *suspended_bdev; |
3ac51e74 DW |
158 | |
159 | /* forced geometry settings */ | |
160 | struct hd_geometry geometry; | |
784aae73 MB |
161 | |
162 | /* sysfs handle */ | |
163 | struct kobject kobj; | |
1da177e4 LT |
164 | }; |
165 | ||
166 | #define MIN_IOS 256 | |
e18b890b CL |
167 | static struct kmem_cache *_io_cache; |
168 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
169 | static struct kmem_cache *_rq_tio_cache; |
170 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 171 | |
1da177e4 LT |
172 | static int __init local_init(void) |
173 | { | |
51157b4a | 174 | int r = -ENOMEM; |
1da177e4 | 175 | |
1da177e4 | 176 | /* allocate a slab for the dm_ios */ |
028867ac | 177 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 178 | if (!_io_cache) |
51157b4a | 179 | return r; |
1da177e4 LT |
180 | |
181 | /* allocate a slab for the target ios */ | |
028867ac | 182 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
183 | if (!_tio_cache) |
184 | goto out_free_io_cache; | |
1da177e4 | 185 | |
8fbf26ad KU |
186 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
187 | if (!_rq_tio_cache) | |
188 | goto out_free_tio_cache; | |
189 | ||
190 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
191 | if (!_rq_bio_info_cache) | |
192 | goto out_free_rq_tio_cache; | |
193 | ||
51e5b2bd | 194 | r = dm_uevent_init(); |
51157b4a | 195 | if (r) |
8fbf26ad | 196 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 197 | |
1da177e4 LT |
198 | _major = major; |
199 | r = register_blkdev(_major, _name); | |
51157b4a KU |
200 | if (r < 0) |
201 | goto out_uevent_exit; | |
1da177e4 LT |
202 | |
203 | if (!_major) | |
204 | _major = r; | |
205 | ||
206 | return 0; | |
51157b4a KU |
207 | |
208 | out_uevent_exit: | |
209 | dm_uevent_exit(); | |
8fbf26ad KU |
210 | out_free_rq_bio_info_cache: |
211 | kmem_cache_destroy(_rq_bio_info_cache); | |
212 | out_free_rq_tio_cache: | |
213 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
214 | out_free_tio_cache: |
215 | kmem_cache_destroy(_tio_cache); | |
216 | out_free_io_cache: | |
217 | kmem_cache_destroy(_io_cache); | |
218 | ||
219 | return r; | |
1da177e4 LT |
220 | } |
221 | ||
222 | static void local_exit(void) | |
223 | { | |
8fbf26ad KU |
224 | kmem_cache_destroy(_rq_bio_info_cache); |
225 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
226 | kmem_cache_destroy(_tio_cache); |
227 | kmem_cache_destroy(_io_cache); | |
00d59405 | 228 | unregister_blkdev(_major, _name); |
51e5b2bd | 229 | dm_uevent_exit(); |
1da177e4 LT |
230 | |
231 | _major = 0; | |
232 | ||
233 | DMINFO("cleaned up"); | |
234 | } | |
235 | ||
b9249e55 | 236 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
237 | local_init, |
238 | dm_target_init, | |
239 | dm_linear_init, | |
240 | dm_stripe_init, | |
945fa4d2 | 241 | dm_kcopyd_init, |
1da177e4 LT |
242 | dm_interface_init, |
243 | }; | |
244 | ||
b9249e55 | 245 | static void (*_exits[])(void) = { |
1da177e4 LT |
246 | local_exit, |
247 | dm_target_exit, | |
248 | dm_linear_exit, | |
249 | dm_stripe_exit, | |
945fa4d2 | 250 | dm_kcopyd_exit, |
1da177e4 LT |
251 | dm_interface_exit, |
252 | }; | |
253 | ||
254 | static int __init dm_init(void) | |
255 | { | |
256 | const int count = ARRAY_SIZE(_inits); | |
257 | ||
258 | int r, i; | |
259 | ||
260 | for (i = 0; i < count; i++) { | |
261 | r = _inits[i](); | |
262 | if (r) | |
263 | goto bad; | |
264 | } | |
265 | ||
266 | return 0; | |
267 | ||
268 | bad: | |
269 | while (i--) | |
270 | _exits[i](); | |
271 | ||
272 | return r; | |
273 | } | |
274 | ||
275 | static void __exit dm_exit(void) | |
276 | { | |
277 | int i = ARRAY_SIZE(_exits); | |
278 | ||
279 | while (i--) | |
280 | _exits[i](); | |
281 | } | |
282 | ||
283 | /* | |
284 | * Block device functions | |
285 | */ | |
fe5f9f2c | 286 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
287 | { |
288 | struct mapped_device *md; | |
289 | ||
fba9f90e JM |
290 | spin_lock(&_minor_lock); |
291 | ||
fe5f9f2c | 292 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
293 | if (!md) |
294 | goto out; | |
295 | ||
5c6bd75d AK |
296 | if (test_bit(DMF_FREEING, &md->flags) || |
297 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
298 | md = NULL; |
299 | goto out; | |
300 | } | |
301 | ||
1da177e4 | 302 | dm_get(md); |
5c6bd75d | 303 | atomic_inc(&md->open_count); |
fba9f90e JM |
304 | |
305 | out: | |
306 | spin_unlock(&_minor_lock); | |
307 | ||
308 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
309 | } |
310 | ||
fe5f9f2c | 311 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 312 | { |
fe5f9f2c | 313 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 314 | atomic_dec(&md->open_count); |
1da177e4 LT |
315 | dm_put(md); |
316 | return 0; | |
317 | } | |
318 | ||
5c6bd75d AK |
319 | int dm_open_count(struct mapped_device *md) |
320 | { | |
321 | return atomic_read(&md->open_count); | |
322 | } | |
323 | ||
324 | /* | |
325 | * Guarantees nothing is using the device before it's deleted. | |
326 | */ | |
327 | int dm_lock_for_deletion(struct mapped_device *md) | |
328 | { | |
329 | int r = 0; | |
330 | ||
331 | spin_lock(&_minor_lock); | |
332 | ||
333 | if (dm_open_count(md)) | |
334 | r = -EBUSY; | |
335 | else | |
336 | set_bit(DMF_DELETING, &md->flags); | |
337 | ||
338 | spin_unlock(&_minor_lock); | |
339 | ||
340 | return r; | |
341 | } | |
342 | ||
3ac51e74 DW |
343 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
344 | { | |
345 | struct mapped_device *md = bdev->bd_disk->private_data; | |
346 | ||
347 | return dm_get_geometry(md, geo); | |
348 | } | |
349 | ||
fe5f9f2c | 350 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
351 | unsigned int cmd, unsigned long arg) |
352 | { | |
fe5f9f2c AV |
353 | struct mapped_device *md = bdev->bd_disk->private_data; |
354 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
355 | struct dm_target *tgt; |
356 | int r = -ENOTTY; | |
357 | ||
aa129a22 MB |
358 | if (!map || !dm_table_get_size(map)) |
359 | goto out; | |
360 | ||
361 | /* We only support devices that have a single target */ | |
362 | if (dm_table_get_num_targets(map) != 1) | |
363 | goto out; | |
364 | ||
365 | tgt = dm_table_get_target(map, 0); | |
366 | ||
367 | if (dm_suspended(md)) { | |
368 | r = -EAGAIN; | |
369 | goto out; | |
370 | } | |
371 | ||
372 | if (tgt->type->ioctl) | |
647b3d00 | 373 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
374 | |
375 | out: | |
376 | dm_table_put(map); | |
377 | ||
aa129a22 MB |
378 | return r; |
379 | } | |
380 | ||
028867ac | 381 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
382 | { |
383 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
384 | } | |
385 | ||
028867ac | 386 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
387 | { |
388 | mempool_free(io, md->io_pool); | |
389 | } | |
390 | ||
028867ac | 391 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
392 | { |
393 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
394 | } | |
395 | ||
028867ac | 396 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
397 | { |
398 | mempool_free(tio, md->tio_pool); | |
399 | } | |
400 | ||
3eaf840e JNN |
401 | static void start_io_acct(struct dm_io *io) |
402 | { | |
403 | struct mapped_device *md = io->md; | |
c9959059 | 404 | int cpu; |
3eaf840e JNN |
405 | |
406 | io->start_time = jiffies; | |
407 | ||
074a7aca TH |
408 | cpu = part_stat_lock(); |
409 | part_round_stats(cpu, &dm_disk(md)->part0); | |
410 | part_stat_unlock(); | |
411 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
412 | } |
413 | ||
d221d2e7 | 414 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
415 | { |
416 | struct mapped_device *md = io->md; | |
417 | struct bio *bio = io->bio; | |
418 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 419 | int pending, cpu; |
3eaf840e JNN |
420 | int rw = bio_data_dir(bio); |
421 | ||
074a7aca TH |
422 | cpu = part_stat_lock(); |
423 | part_round_stats(cpu, &dm_disk(md)->part0); | |
424 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
425 | part_stat_unlock(); | |
3eaf840e | 426 | |
074a7aca TH |
427 | dm_disk(md)->part0.in_flight = pending = |
428 | atomic_dec_return(&md->pending); | |
3eaf840e | 429 | |
d221d2e7 MP |
430 | /* nudge anyone waiting on suspend queue */ |
431 | if (!pending) | |
432 | wake_up(&md->wait); | |
3eaf840e JNN |
433 | } |
434 | ||
1da177e4 LT |
435 | /* |
436 | * Add the bio to the list of deferred io. | |
437 | */ | |
438 | static int queue_io(struct mapped_device *md, struct bio *bio) | |
439 | { | |
2ca3310e | 440 | down_write(&md->io_lock); |
1da177e4 LT |
441 | |
442 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 443 | up_write(&md->io_lock); |
1da177e4 LT |
444 | return 1; |
445 | } | |
446 | ||
022c2611 | 447 | spin_lock_irq(&md->deferred_lock); |
1da177e4 | 448 | bio_list_add(&md->deferred, bio); |
022c2611 | 449 | spin_unlock_irq(&md->deferred_lock); |
1da177e4 | 450 | |
2ca3310e | 451 | up_write(&md->io_lock); |
1da177e4 LT |
452 | return 0; /* deferred successfully */ |
453 | } | |
454 | ||
455 | /* | |
456 | * Everyone (including functions in this file), should use this | |
457 | * function to access the md->map field, and make sure they call | |
458 | * dm_table_put() when finished. | |
459 | */ | |
460 | struct dm_table *dm_get_table(struct mapped_device *md) | |
461 | { | |
462 | struct dm_table *t; | |
463 | ||
464 | read_lock(&md->map_lock); | |
465 | t = md->map; | |
466 | if (t) | |
467 | dm_table_get(t); | |
468 | read_unlock(&md->map_lock); | |
469 | ||
470 | return t; | |
471 | } | |
472 | ||
3ac51e74 DW |
473 | /* |
474 | * Get the geometry associated with a dm device | |
475 | */ | |
476 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
477 | { | |
478 | *geo = md->geometry; | |
479 | ||
480 | return 0; | |
481 | } | |
482 | ||
483 | /* | |
484 | * Set the geometry of a device. | |
485 | */ | |
486 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
487 | { | |
488 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
489 | ||
490 | if (geo->start > sz) { | |
491 | DMWARN("Start sector is beyond the geometry limits."); | |
492 | return -EINVAL; | |
493 | } | |
494 | ||
495 | md->geometry = *geo; | |
496 | ||
497 | return 0; | |
498 | } | |
499 | ||
1da177e4 LT |
500 | /*----------------------------------------------------------------- |
501 | * CRUD START: | |
502 | * A more elegant soln is in the works that uses the queue | |
503 | * merge fn, unfortunately there are a couple of changes to | |
504 | * the block layer that I want to make for this. So in the | |
505 | * interests of getting something for people to use I give | |
506 | * you this clearly demarcated crap. | |
507 | *---------------------------------------------------------------*/ | |
508 | ||
2e93ccc1 KU |
509 | static int __noflush_suspending(struct mapped_device *md) |
510 | { | |
511 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
512 | } | |
513 | ||
1da177e4 LT |
514 | /* |
515 | * Decrements the number of outstanding ios that a bio has been | |
516 | * cloned into, completing the original io if necc. | |
517 | */ | |
858119e1 | 518 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 519 | { |
2e93ccc1 | 520 | unsigned long flags; |
b35f8caa MB |
521 | int io_error; |
522 | struct bio *bio; | |
523 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
524 | |
525 | /* Push-back supersedes any I/O errors */ | |
b35f8caa | 526 | if (error && !(io->error > 0 && __noflush_suspending(md))) |
1da177e4 LT |
527 | io->error = error; |
528 | ||
529 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
530 | if (io->error == DM_ENDIO_REQUEUE) { |
531 | /* | |
532 | * Target requested pushing back the I/O. | |
2e93ccc1 | 533 | */ |
022c2611 | 534 | spin_lock_irqsave(&md->deferred_lock, flags); |
b35f8caa | 535 | if (__noflush_suspending(md)) |
022c2611 | 536 | bio_list_add(&md->deferred, io->bio); |
2e93ccc1 KU |
537 | else |
538 | /* noflush suspend was interrupted. */ | |
539 | io->error = -EIO; | |
022c2611 | 540 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
541 | } |
542 | ||
d221d2e7 | 543 | end_io_acct(io); |
1da177e4 | 544 | |
b35f8caa MB |
545 | io_error = io->error; |
546 | bio = io->bio; | |
2e93ccc1 | 547 | |
b35f8caa MB |
548 | free_io(md, io); |
549 | ||
550 | if (io_error != DM_ENDIO_REQUEUE) { | |
551 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 552 | |
b35f8caa MB |
553 | bio_endio(bio, io_error); |
554 | } | |
1da177e4 LT |
555 | } |
556 | } | |
557 | ||
6712ecf8 | 558 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
559 | { |
560 | int r = 0; | |
028867ac | 561 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 562 | struct dm_io *io = tio->io; |
9faf400f | 563 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
564 | dm_endio_fn endio = tio->ti->type->end_io; |
565 | ||
1da177e4 LT |
566 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
567 | error = -EIO; | |
568 | ||
569 | if (endio) { | |
570 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
571 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
572 | /* | |
573 | * error and requeue request are handled | |
574 | * in dec_pending(). | |
575 | */ | |
1da177e4 | 576 | error = r; |
45cbcd79 KU |
577 | else if (r == DM_ENDIO_INCOMPLETE) |
578 | /* The target will handle the io */ | |
6712ecf8 | 579 | return; |
45cbcd79 KU |
580 | else if (r) { |
581 | DMWARN("unimplemented target endio return value: %d", r); | |
582 | BUG(); | |
583 | } | |
1da177e4 LT |
584 | } |
585 | ||
9faf400f SB |
586 | /* |
587 | * Store md for cleanup instead of tio which is about to get freed. | |
588 | */ | |
589 | bio->bi_private = md->bs; | |
590 | ||
9faf400f | 591 | free_tio(md, tio); |
b35f8caa MB |
592 | bio_put(bio); |
593 | dec_pending(io, error); | |
1da177e4 LT |
594 | } |
595 | ||
596 | static sector_t max_io_len(struct mapped_device *md, | |
597 | sector_t sector, struct dm_target *ti) | |
598 | { | |
599 | sector_t offset = sector - ti->begin; | |
600 | sector_t len = ti->len - offset; | |
601 | ||
602 | /* | |
603 | * Does the target need to split even further ? | |
604 | */ | |
605 | if (ti->split_io) { | |
606 | sector_t boundary; | |
607 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
608 | - offset; | |
609 | if (len > boundary) | |
610 | len = boundary; | |
611 | } | |
612 | ||
613 | return len; | |
614 | } | |
615 | ||
616 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 617 | struct dm_target_io *tio) |
1da177e4 LT |
618 | { |
619 | int r; | |
2056a782 | 620 | sector_t sector; |
9faf400f | 621 | struct mapped_device *md; |
1da177e4 LT |
622 | |
623 | /* | |
624 | * Sanity checks. | |
625 | */ | |
626 | BUG_ON(!clone->bi_size); | |
627 | ||
628 | clone->bi_end_io = clone_endio; | |
629 | clone->bi_private = tio; | |
630 | ||
631 | /* | |
632 | * Map the clone. If r == 0 we don't need to do | |
633 | * anything, the target has assumed ownership of | |
634 | * this io. | |
635 | */ | |
636 | atomic_inc(&tio->io->io_count); | |
2056a782 | 637 | sector = clone->bi_sector; |
1da177e4 | 638 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 639 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 640 | /* the bio has been remapped so dispatch it */ |
2056a782 | 641 | |
5f3ea37c | 642 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
643 | tio->io->bio->bi_bdev->bd_dev, |
644 | clone->bi_sector, sector); | |
2056a782 | 645 | |
1da177e4 | 646 | generic_make_request(clone); |
2e93ccc1 KU |
647 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
648 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
649 | md = tio->io->md; |
650 | dec_pending(tio->io, r); | |
651 | /* | |
652 | * Store bio_set for cleanup. | |
653 | */ | |
654 | clone->bi_private = md->bs; | |
1da177e4 | 655 | bio_put(clone); |
9faf400f | 656 | free_tio(md, tio); |
45cbcd79 KU |
657 | } else if (r) { |
658 | DMWARN("unimplemented target map return value: %d", r); | |
659 | BUG(); | |
1da177e4 LT |
660 | } |
661 | } | |
662 | ||
663 | struct clone_info { | |
664 | struct mapped_device *md; | |
665 | struct dm_table *map; | |
666 | struct bio *bio; | |
667 | struct dm_io *io; | |
668 | sector_t sector; | |
669 | sector_t sector_count; | |
670 | unsigned short idx; | |
671 | }; | |
672 | ||
3676347a PO |
673 | static void dm_bio_destructor(struct bio *bio) |
674 | { | |
9faf400f SB |
675 | struct bio_set *bs = bio->bi_private; |
676 | ||
677 | bio_free(bio, bs); | |
3676347a PO |
678 | } |
679 | ||
1da177e4 LT |
680 | /* |
681 | * Creates a little bio that is just does part of a bvec. | |
682 | */ | |
683 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
684 | unsigned short idx, unsigned int offset, | |
9faf400f | 685 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
686 | { |
687 | struct bio *clone; | |
688 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
689 | ||
9faf400f | 690 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 691 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
692 | *clone->bi_io_vec = *bv; |
693 | ||
694 | clone->bi_sector = sector; | |
695 | clone->bi_bdev = bio->bi_bdev; | |
696 | clone->bi_rw = bio->bi_rw; | |
697 | clone->bi_vcnt = 1; | |
698 | clone->bi_size = to_bytes(len); | |
699 | clone->bi_io_vec->bv_offset = offset; | |
700 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 701 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 702 | |
9c47008d MP |
703 | if (bio_integrity(bio)) { |
704 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
705 | bio_integrity_trim(clone, | |
706 | bio_sector_offset(bio, idx, offset), len); | |
707 | } | |
708 | ||
1da177e4 LT |
709 | return clone; |
710 | } | |
711 | ||
712 | /* | |
713 | * Creates a bio that consists of range of complete bvecs. | |
714 | */ | |
715 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
716 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 717 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
718 | { |
719 | struct bio *clone; | |
720 | ||
9faf400f SB |
721 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
722 | __bio_clone(clone, bio); | |
723 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
724 | clone->bi_sector = sector; |
725 | clone->bi_idx = idx; | |
726 | clone->bi_vcnt = idx + bv_count; | |
727 | clone->bi_size = to_bytes(len); | |
728 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
729 | ||
9c47008d MP |
730 | if (bio_integrity(bio)) { |
731 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
732 | ||
733 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
734 | bio_integrity_trim(clone, | |
735 | bio_sector_offset(bio, idx, 0), len); | |
736 | } | |
737 | ||
1da177e4 LT |
738 | return clone; |
739 | } | |
740 | ||
512875bd | 741 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
742 | { |
743 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
744 | struct dm_target *ti; |
745 | sector_t len = 0, max; | |
028867ac | 746 | struct dm_target_io *tio; |
1da177e4 | 747 | |
512875bd JN |
748 | ti = dm_table_find_target(ci->map, ci->sector); |
749 | if (!dm_target_is_valid(ti)) | |
750 | return -EIO; | |
751 | ||
752 | max = max_io_len(ci->md, ci->sector, ti); | |
753 | ||
1da177e4 LT |
754 | /* |
755 | * Allocate a target io object. | |
756 | */ | |
757 | tio = alloc_tio(ci->md); | |
758 | tio->io = ci->io; | |
759 | tio->ti = ti; | |
760 | memset(&tio->info, 0, sizeof(tio->info)); | |
761 | ||
762 | if (ci->sector_count <= max) { | |
763 | /* | |
764 | * Optimise for the simple case where we can do all of | |
765 | * the remaining io with a single clone. | |
766 | */ | |
767 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
768 | bio->bi_vcnt - ci->idx, ci->sector_count, |
769 | ci->md->bs); | |
1da177e4 LT |
770 | __map_bio(ti, clone, tio); |
771 | ci->sector_count = 0; | |
772 | ||
773 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
774 | /* | |
775 | * There are some bvecs that don't span targets. | |
776 | * Do as many of these as possible. | |
777 | */ | |
778 | int i; | |
779 | sector_t remaining = max; | |
780 | sector_t bv_len; | |
781 | ||
782 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
783 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
784 | ||
785 | if (bv_len > remaining) | |
786 | break; | |
787 | ||
788 | remaining -= bv_len; | |
789 | len += bv_len; | |
790 | } | |
791 | ||
9faf400f SB |
792 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
793 | ci->md->bs); | |
1da177e4 LT |
794 | __map_bio(ti, clone, tio); |
795 | ||
796 | ci->sector += len; | |
797 | ci->sector_count -= len; | |
798 | ci->idx = i; | |
799 | ||
800 | } else { | |
801 | /* | |
d2044a94 | 802 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
803 | */ |
804 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
805 | sector_t remaining = to_sector(bv->bv_len); |
806 | unsigned int offset = 0; | |
1da177e4 | 807 | |
d2044a94 AK |
808 | do { |
809 | if (offset) { | |
810 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
811 | if (!dm_target_is_valid(ti)) |
812 | return -EIO; | |
813 | ||
d2044a94 | 814 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 815 | |
d2044a94 AK |
816 | tio = alloc_tio(ci->md); |
817 | tio->io = ci->io; | |
818 | tio->ti = ti; | |
819 | memset(&tio->info, 0, sizeof(tio->info)); | |
820 | } | |
821 | ||
822 | len = min(remaining, max); | |
823 | ||
824 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
825 | bv->bv_offset + offset, len, |
826 | ci->md->bs); | |
d2044a94 AK |
827 | |
828 | __map_bio(ti, clone, tio); | |
829 | ||
830 | ci->sector += len; | |
831 | ci->sector_count -= len; | |
832 | offset += to_bytes(len); | |
833 | } while (remaining -= len); | |
1da177e4 | 834 | |
1da177e4 LT |
835 | ci->idx++; |
836 | } | |
512875bd JN |
837 | |
838 | return 0; | |
1da177e4 LT |
839 | } |
840 | ||
841 | /* | |
8a53c28d | 842 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 843 | */ |
f0b9a450 | 844 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
845 | { |
846 | struct clone_info ci; | |
512875bd | 847 | int error = 0; |
1da177e4 LT |
848 | |
849 | ci.map = dm_get_table(md); | |
f0b9a450 MP |
850 | if (unlikely(!ci.map)) { |
851 | bio_io_error(bio); | |
852 | return; | |
853 | } | |
ab4c1424 AK |
854 | if (unlikely(bio_barrier(bio) && !dm_table_barrier_ok(ci.map))) { |
855 | dm_table_put(ci.map); | |
856 | bio_endio(bio, -EOPNOTSUPP); | |
f0b9a450 | 857 | return; |
ab4c1424 | 858 | } |
1da177e4 LT |
859 | ci.md = md; |
860 | ci.bio = bio; | |
861 | ci.io = alloc_io(md); | |
862 | ci.io->error = 0; | |
863 | atomic_set(&ci.io->io_count, 1); | |
864 | ci.io->bio = bio; | |
865 | ci.io->md = md; | |
866 | ci.sector = bio->bi_sector; | |
867 | ci.sector_count = bio_sectors(bio); | |
868 | ci.idx = bio->bi_idx; | |
869 | ||
3eaf840e | 870 | start_io_acct(ci.io); |
512875bd JN |
871 | while (ci.sector_count && !error) |
872 | error = __clone_and_map(&ci); | |
1da177e4 LT |
873 | |
874 | /* drop the extra reference count */ | |
512875bd | 875 | dec_pending(ci.io, error); |
1da177e4 LT |
876 | dm_table_put(ci.map); |
877 | } | |
878 | /*----------------------------------------------------------------- | |
879 | * CRUD END | |
880 | *---------------------------------------------------------------*/ | |
881 | ||
f6fccb12 MB |
882 | static int dm_merge_bvec(struct request_queue *q, |
883 | struct bvec_merge_data *bvm, | |
884 | struct bio_vec *biovec) | |
885 | { | |
886 | struct mapped_device *md = q->queuedata; | |
887 | struct dm_table *map = dm_get_table(md); | |
888 | struct dm_target *ti; | |
889 | sector_t max_sectors; | |
5037108a | 890 | int max_size = 0; |
f6fccb12 MB |
891 | |
892 | if (unlikely(!map)) | |
5037108a | 893 | goto out; |
f6fccb12 MB |
894 | |
895 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
896 | if (!dm_target_is_valid(ti)) |
897 | goto out_table; | |
f6fccb12 MB |
898 | |
899 | /* | |
900 | * Find maximum amount of I/O that won't need splitting | |
901 | */ | |
902 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
903 | (sector_t) BIO_MAX_SECTORS); | |
904 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
905 | if (max_size < 0) | |
906 | max_size = 0; | |
907 | ||
908 | /* | |
909 | * merge_bvec_fn() returns number of bytes | |
910 | * it can accept at this offset | |
911 | * max is precomputed maximal io size | |
912 | */ | |
913 | if (max_size && ti->type->merge) | |
914 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
915 | ||
b01cd5ac | 916 | out_table: |
5037108a MP |
917 | dm_table_put(map); |
918 | ||
919 | out: | |
f6fccb12 MB |
920 | /* |
921 | * Always allow an entire first page | |
922 | */ | |
923 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
924 | max_size = biovec->bv_len; | |
925 | ||
f6fccb12 MB |
926 | return max_size; |
927 | } | |
928 | ||
1da177e4 LT |
929 | /* |
930 | * The request function that just remaps the bio built up by | |
931 | * dm_merge_bvec. | |
932 | */ | |
165125e1 | 933 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 934 | { |
9e4e5f87 | 935 | int r = -EIO; |
12f03a49 | 936 | int rw = bio_data_dir(bio); |
1da177e4 | 937 | struct mapped_device *md = q->queuedata; |
c9959059 | 938 | int cpu; |
1da177e4 | 939 | |
2ca3310e | 940 | down_read(&md->io_lock); |
1da177e4 | 941 | |
074a7aca TH |
942 | cpu = part_stat_lock(); |
943 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
944 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
945 | part_stat_unlock(); | |
12f03a49 | 946 | |
1da177e4 LT |
947 | /* |
948 | * If we're suspended we have to queue | |
949 | * this io for later. | |
950 | */ | |
951 | while (test_bit(DMF_BLOCK_IO, &md->flags)) { | |
2ca3310e | 952 | up_read(&md->io_lock); |
1da177e4 | 953 | |
9e4e5f87 MB |
954 | if (bio_rw(bio) != READA) |
955 | r = queue_io(md, bio); | |
1da177e4 | 956 | |
9e4e5f87 MB |
957 | if (r <= 0) |
958 | goto out_req; | |
1da177e4 LT |
959 | |
960 | /* | |
961 | * We're in a while loop, because someone could suspend | |
962 | * before we get to the following read lock. | |
963 | */ | |
2ca3310e | 964 | down_read(&md->io_lock); |
1da177e4 LT |
965 | } |
966 | ||
f0b9a450 | 967 | __split_and_process_bio(md, bio); |
2ca3310e | 968 | up_read(&md->io_lock); |
f0b9a450 | 969 | return 0; |
9e4e5f87 MB |
970 | |
971 | out_req: | |
972 | if (r < 0) | |
973 | bio_io_error(bio); | |
974 | ||
1da177e4 LT |
975 | return 0; |
976 | } | |
977 | ||
165125e1 | 978 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
979 | { |
980 | struct mapped_device *md = q->queuedata; | |
981 | struct dm_table *map = dm_get_table(md); | |
982 | ||
983 | if (map) { | |
984 | dm_table_unplug_all(map); | |
985 | dm_table_put(map); | |
986 | } | |
987 | } | |
988 | ||
989 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
990 | { | |
8a57dfc6 CS |
991 | int r = bdi_bits; |
992 | struct mapped_device *md = congested_data; | |
993 | struct dm_table *map; | |
1da177e4 | 994 | |
8a57dfc6 CS |
995 | if (!test_bit(DMF_BLOCK_IO, &md->flags)) { |
996 | map = dm_get_table(md); | |
997 | if (map) { | |
998 | r = dm_table_any_congested(map, bdi_bits); | |
999 | dm_table_put(map); | |
1000 | } | |
1001 | } | |
1002 | ||
1da177e4 LT |
1003 | return r; |
1004 | } | |
1005 | ||
1006 | /*----------------------------------------------------------------- | |
1007 | * An IDR is used to keep track of allocated minor numbers. | |
1008 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1009 | static DEFINE_IDR(_minor_idr); |
1010 | ||
2b06cfff | 1011 | static void free_minor(int minor) |
1da177e4 | 1012 | { |
f32c10b0 | 1013 | spin_lock(&_minor_lock); |
1da177e4 | 1014 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1015 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1016 | } |
1017 | ||
1018 | /* | |
1019 | * See if the device with a specific minor # is free. | |
1020 | */ | |
cf13ab8e | 1021 | static int specific_minor(int minor) |
1da177e4 LT |
1022 | { |
1023 | int r, m; | |
1024 | ||
1025 | if (minor >= (1 << MINORBITS)) | |
1026 | return -EINVAL; | |
1027 | ||
62f75c2f JM |
1028 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1029 | if (!r) | |
1030 | return -ENOMEM; | |
1031 | ||
f32c10b0 | 1032 | spin_lock(&_minor_lock); |
1da177e4 LT |
1033 | |
1034 | if (idr_find(&_minor_idr, minor)) { | |
1035 | r = -EBUSY; | |
1036 | goto out; | |
1037 | } | |
1038 | ||
ba61fdd1 | 1039 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1040 | if (r) |
1da177e4 | 1041 | goto out; |
1da177e4 LT |
1042 | |
1043 | if (m != minor) { | |
1044 | idr_remove(&_minor_idr, m); | |
1045 | r = -EBUSY; | |
1046 | goto out; | |
1047 | } | |
1048 | ||
1049 | out: | |
f32c10b0 | 1050 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1051 | return r; |
1052 | } | |
1053 | ||
cf13ab8e | 1054 | static int next_free_minor(int *minor) |
1da177e4 | 1055 | { |
2b06cfff | 1056 | int r, m; |
1da177e4 | 1057 | |
1da177e4 | 1058 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1059 | if (!r) |
1060 | return -ENOMEM; | |
1061 | ||
f32c10b0 | 1062 | spin_lock(&_minor_lock); |
1da177e4 | 1063 | |
ba61fdd1 | 1064 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1065 | if (r) |
1da177e4 | 1066 | goto out; |
1da177e4 LT |
1067 | |
1068 | if (m >= (1 << MINORBITS)) { | |
1069 | idr_remove(&_minor_idr, m); | |
1070 | r = -ENOSPC; | |
1071 | goto out; | |
1072 | } | |
1073 | ||
1074 | *minor = m; | |
1075 | ||
1076 | out: | |
f32c10b0 | 1077 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1078 | return r; |
1079 | } | |
1080 | ||
1081 | static struct block_device_operations dm_blk_dops; | |
1082 | ||
53d5914f MP |
1083 | static void dm_wq_work(struct work_struct *work); |
1084 | ||
1da177e4 LT |
1085 | /* |
1086 | * Allocate and initialise a blank device with a given minor. | |
1087 | */ | |
2b06cfff | 1088 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1089 | { |
1090 | int r; | |
cf13ab8e | 1091 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1092 | void *old_md; |
1da177e4 LT |
1093 | |
1094 | if (!md) { | |
1095 | DMWARN("unable to allocate device, out of memory."); | |
1096 | return NULL; | |
1097 | } | |
1098 | ||
10da4f79 | 1099 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1100 | goto bad_module_get; |
10da4f79 | 1101 | |
1da177e4 | 1102 | /* get a minor number for the dev */ |
2b06cfff | 1103 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1104 | r = next_free_minor(&minor); |
2b06cfff | 1105 | else |
cf13ab8e | 1106 | r = specific_minor(minor); |
1da177e4 | 1107 | if (r < 0) |
6ed7ade8 | 1108 | goto bad_minor; |
1da177e4 | 1109 | |
2ca3310e | 1110 | init_rwsem(&md->io_lock); |
e61290a4 | 1111 | mutex_init(&md->suspend_lock); |
022c2611 | 1112 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1113 | rwlock_init(&md->map_lock); |
1114 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1115 | atomic_set(&md->open_count, 0); |
1da177e4 | 1116 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1117 | atomic_set(&md->uevent_seq, 0); |
1118 | INIT_LIST_HEAD(&md->uevent_list); | |
1119 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1120 | |
1121 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1122 | if (!md->queue) | |
6ed7ade8 | 1123 | goto bad_queue; |
1da177e4 LT |
1124 | |
1125 | md->queue->queuedata = md; | |
1126 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1127 | md->queue->backing_dev_info.congested_data = md; | |
1128 | blk_queue_make_request(md->queue, dm_request); | |
99360b4c | 1129 | blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN, NULL); |
daef265f | 1130 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1131 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1132 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1133 | |
93d2341c | 1134 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1135 | if (!md->io_pool) |
6ed7ade8 | 1136 | goto bad_io_pool; |
1da177e4 | 1137 | |
93d2341c | 1138 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1139 | if (!md->tio_pool) |
6ed7ade8 | 1140 | goto bad_tio_pool; |
1da177e4 | 1141 | |
bb799ca0 | 1142 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1143 | if (!md->bs) |
1144 | goto bad_no_bioset; | |
1145 | ||
1da177e4 LT |
1146 | md->disk = alloc_disk(1); |
1147 | if (!md->disk) | |
6ed7ade8 | 1148 | goto bad_disk; |
1da177e4 | 1149 | |
f0b04115 JM |
1150 | atomic_set(&md->pending, 0); |
1151 | init_waitqueue_head(&md->wait); | |
53d5914f | 1152 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1153 | init_waitqueue_head(&md->eventq); |
1154 | ||
1da177e4 LT |
1155 | md->disk->major = _major; |
1156 | md->disk->first_minor = minor; | |
1157 | md->disk->fops = &dm_blk_dops; | |
1158 | md->disk->queue = md->queue; | |
1159 | md->disk->private_data = md; | |
1160 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1161 | add_disk(md->disk); | |
7e51f257 | 1162 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1163 | |
304f3f6a MB |
1164 | md->wq = create_singlethread_workqueue("kdmflush"); |
1165 | if (!md->wq) | |
1166 | goto bad_thread; | |
1167 | ||
ba61fdd1 | 1168 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1169 | spin_lock(&_minor_lock); |
ba61fdd1 | 1170 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1171 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1172 | |
1173 | BUG_ON(old_md != MINOR_ALLOCED); | |
1174 | ||
1da177e4 LT |
1175 | return md; |
1176 | ||
304f3f6a MB |
1177 | bad_thread: |
1178 | put_disk(md->disk); | |
6ed7ade8 | 1179 | bad_disk: |
9faf400f | 1180 | bioset_free(md->bs); |
6ed7ade8 | 1181 | bad_no_bioset: |
1da177e4 | 1182 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1183 | bad_tio_pool: |
1da177e4 | 1184 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1185 | bad_io_pool: |
1312f40e | 1186 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1187 | bad_queue: |
1da177e4 | 1188 | free_minor(minor); |
6ed7ade8 | 1189 | bad_minor: |
10da4f79 | 1190 | module_put(THIS_MODULE); |
6ed7ade8 | 1191 | bad_module_get: |
1da177e4 LT |
1192 | kfree(md); |
1193 | return NULL; | |
1194 | } | |
1195 | ||
ae9da83f JN |
1196 | static void unlock_fs(struct mapped_device *md); |
1197 | ||
1da177e4 LT |
1198 | static void free_dev(struct mapped_device *md) |
1199 | { | |
f331c029 | 1200 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1201 | |
d9dde59b | 1202 | if (md->suspended_bdev) { |
ae9da83f | 1203 | unlock_fs(md); |
d9dde59b JN |
1204 | bdput(md->suspended_bdev); |
1205 | } | |
304f3f6a | 1206 | destroy_workqueue(md->wq); |
1da177e4 LT |
1207 | mempool_destroy(md->tio_pool); |
1208 | mempool_destroy(md->io_pool); | |
9faf400f | 1209 | bioset_free(md->bs); |
9c47008d | 1210 | blk_integrity_unregister(md->disk); |
1da177e4 | 1211 | del_gendisk(md->disk); |
63d94e48 | 1212 | free_minor(minor); |
fba9f90e JM |
1213 | |
1214 | spin_lock(&_minor_lock); | |
1215 | md->disk->private_data = NULL; | |
1216 | spin_unlock(&_minor_lock); | |
1217 | ||
1da177e4 | 1218 | put_disk(md->disk); |
1312f40e | 1219 | blk_cleanup_queue(md->queue); |
10da4f79 | 1220 | module_put(THIS_MODULE); |
1da177e4 LT |
1221 | kfree(md); |
1222 | } | |
1223 | ||
1224 | /* | |
1225 | * Bind a table to the device. | |
1226 | */ | |
1227 | static void event_callback(void *context) | |
1228 | { | |
7a8c3d3b MA |
1229 | unsigned long flags; |
1230 | LIST_HEAD(uevents); | |
1da177e4 LT |
1231 | struct mapped_device *md = (struct mapped_device *) context; |
1232 | ||
7a8c3d3b MA |
1233 | spin_lock_irqsave(&md->uevent_lock, flags); |
1234 | list_splice_init(&md->uevent_list, &uevents); | |
1235 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1236 | ||
ed9e1982 | 1237 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1238 | |
1da177e4 LT |
1239 | atomic_inc(&md->event_nr); |
1240 | wake_up(&md->eventq); | |
1241 | } | |
1242 | ||
4e90188b | 1243 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1244 | { |
4e90188b | 1245 | set_capacity(md->disk, size); |
1da177e4 | 1246 | |
1b1dcc1b | 1247 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1248 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1249 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1250 | } |
1251 | ||
1252 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1253 | { | |
165125e1 | 1254 | struct request_queue *q = md->queue; |
1da177e4 LT |
1255 | sector_t size; |
1256 | ||
1257 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1258 | |
1259 | /* | |
1260 | * Wipe any geometry if the size of the table changed. | |
1261 | */ | |
1262 | if (size != get_capacity(md->disk)) | |
1263 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1264 | ||
bfa152fa JN |
1265 | if (md->suspended_bdev) |
1266 | __set_size(md, size); | |
d5816876 MP |
1267 | |
1268 | if (!size) { | |
1269 | dm_table_destroy(t); | |
1da177e4 | 1270 | return 0; |
d5816876 | 1271 | } |
1da177e4 | 1272 | |
2ca3310e AK |
1273 | dm_table_event_callback(t, event_callback, md); |
1274 | ||
1da177e4 LT |
1275 | write_lock(&md->map_lock); |
1276 | md->map = t; | |
2ca3310e | 1277 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1278 | write_unlock(&md->map_lock); |
1279 | ||
1da177e4 LT |
1280 | return 0; |
1281 | } | |
1282 | ||
1283 | static void __unbind(struct mapped_device *md) | |
1284 | { | |
1285 | struct dm_table *map = md->map; | |
1286 | ||
1287 | if (!map) | |
1288 | return; | |
1289 | ||
1290 | dm_table_event_callback(map, NULL, NULL); | |
1291 | write_lock(&md->map_lock); | |
1292 | md->map = NULL; | |
1293 | write_unlock(&md->map_lock); | |
d5816876 | 1294 | dm_table_destroy(map); |
1da177e4 LT |
1295 | } |
1296 | ||
1297 | /* | |
1298 | * Constructor for a new device. | |
1299 | */ | |
2b06cfff | 1300 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1301 | { |
1302 | struct mapped_device *md; | |
1303 | ||
2b06cfff | 1304 | md = alloc_dev(minor); |
1da177e4 LT |
1305 | if (!md) |
1306 | return -ENXIO; | |
1307 | ||
784aae73 MB |
1308 | dm_sysfs_init(md); |
1309 | ||
1da177e4 LT |
1310 | *result = md; |
1311 | return 0; | |
1312 | } | |
1313 | ||
637842cf | 1314 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1315 | { |
1316 | struct mapped_device *md; | |
1da177e4 LT |
1317 | unsigned minor = MINOR(dev); |
1318 | ||
1319 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1320 | return NULL; | |
1321 | ||
f32c10b0 | 1322 | spin_lock(&_minor_lock); |
1da177e4 LT |
1323 | |
1324 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1325 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1326 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1327 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1328 | md = NULL; |
fba9f90e JM |
1329 | goto out; |
1330 | } | |
1da177e4 | 1331 | |
fba9f90e | 1332 | out: |
f32c10b0 | 1333 | spin_unlock(&_minor_lock); |
1da177e4 | 1334 | |
637842cf DT |
1335 | return md; |
1336 | } | |
1337 | ||
d229a958 DT |
1338 | struct mapped_device *dm_get_md(dev_t dev) |
1339 | { | |
1340 | struct mapped_device *md = dm_find_md(dev); | |
1341 | ||
1342 | if (md) | |
1343 | dm_get(md); | |
1344 | ||
1345 | return md; | |
1346 | } | |
1347 | ||
9ade92a9 | 1348 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1349 | { |
9ade92a9 | 1350 | return md->interface_ptr; |
1da177e4 LT |
1351 | } |
1352 | ||
1353 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1354 | { | |
1355 | md->interface_ptr = ptr; | |
1356 | } | |
1357 | ||
1358 | void dm_get(struct mapped_device *md) | |
1359 | { | |
1360 | atomic_inc(&md->holders); | |
1361 | } | |
1362 | ||
72d94861 AK |
1363 | const char *dm_device_name(struct mapped_device *md) |
1364 | { | |
1365 | return md->name; | |
1366 | } | |
1367 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1368 | ||
1da177e4 LT |
1369 | void dm_put(struct mapped_device *md) |
1370 | { | |
1134e5ae | 1371 | struct dm_table *map; |
1da177e4 | 1372 | |
fba9f90e JM |
1373 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1374 | ||
f32c10b0 | 1375 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1376 | map = dm_get_table(md); |
f331c029 TH |
1377 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1378 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1379 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1380 | spin_unlock(&_minor_lock); |
cf222b37 | 1381 | if (!dm_suspended(md)) { |
1da177e4 LT |
1382 | dm_table_presuspend_targets(map); |
1383 | dm_table_postsuspend_targets(map); | |
1384 | } | |
784aae73 | 1385 | dm_sysfs_exit(md); |
1134e5ae | 1386 | dm_table_put(map); |
a1b51e98 | 1387 | __unbind(md); |
1da177e4 LT |
1388 | free_dev(md); |
1389 | } | |
1da177e4 | 1390 | } |
79eb885c | 1391 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1392 | |
401600df | 1393 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
1394 | { |
1395 | int r = 0; | |
b44ebeb0 MP |
1396 | DECLARE_WAITQUEUE(wait, current); |
1397 | ||
1398 | dm_unplug_all(md->queue); | |
1399 | ||
1400 | add_wait_queue(&md->wait, &wait); | |
46125c1c MB |
1401 | |
1402 | while (1) { | |
401600df | 1403 | set_current_state(interruptible); |
46125c1c MB |
1404 | |
1405 | smp_mb(); | |
1406 | if (!atomic_read(&md->pending)) | |
1407 | break; | |
1408 | ||
401600df MP |
1409 | if (interruptible == TASK_INTERRUPTIBLE && |
1410 | signal_pending(current)) { | |
46125c1c MB |
1411 | r = -EINTR; |
1412 | break; | |
1413 | } | |
1414 | ||
1415 | io_schedule(); | |
1416 | } | |
1417 | set_current_state(TASK_RUNNING); | |
1418 | ||
b44ebeb0 MP |
1419 | remove_wait_queue(&md->wait, &wait); |
1420 | ||
46125c1c MB |
1421 | return r; |
1422 | } | |
1423 | ||
1da177e4 LT |
1424 | /* |
1425 | * Process the deferred bios | |
1426 | */ | |
ef208587 | 1427 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 1428 | { |
ef208587 MP |
1429 | struct mapped_device *md = container_of(work, struct mapped_device, |
1430 | work); | |
6d6f10df | 1431 | struct bio *c; |
1da177e4 | 1432 | |
ef208587 MP |
1433 | down_write(&md->io_lock); |
1434 | ||
022c2611 MP |
1435 | next_bio: |
1436 | spin_lock_irq(&md->deferred_lock); | |
1437 | c = bio_list_pop(&md->deferred); | |
1438 | spin_unlock_irq(&md->deferred_lock); | |
1439 | ||
1440 | if (c) { | |
f0b9a450 | 1441 | __split_and_process_bio(md, c); |
022c2611 MP |
1442 | goto next_bio; |
1443 | } | |
73d410c0 MB |
1444 | |
1445 | clear_bit(DMF_BLOCK_IO, &md->flags); | |
ef208587 MP |
1446 | |
1447 | up_write(&md->io_lock); | |
1da177e4 LT |
1448 | } |
1449 | ||
9a1fb464 | 1450 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 1451 | { |
53d5914f | 1452 | queue_work(md->wq, &md->work); |
304f3f6a MB |
1453 | flush_workqueue(md->wq); |
1454 | } | |
1455 | ||
1da177e4 LT |
1456 | /* |
1457 | * Swap in a new table (destroying old one). | |
1458 | */ | |
1459 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1460 | { | |
93c534ae | 1461 | int r = -EINVAL; |
1da177e4 | 1462 | |
e61290a4 | 1463 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1464 | |
1465 | /* device must be suspended */ | |
cf222b37 | 1466 | if (!dm_suspended(md)) |
93c534ae | 1467 | goto out; |
1da177e4 | 1468 | |
bfa152fa JN |
1469 | /* without bdev, the device size cannot be changed */ |
1470 | if (!md->suspended_bdev) | |
1471 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1472 | goto out; | |
1473 | ||
1da177e4 LT |
1474 | __unbind(md); |
1475 | r = __bind(md, table); | |
1da177e4 | 1476 | |
93c534ae | 1477 | out: |
e61290a4 | 1478 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1479 | return r; |
1da177e4 LT |
1480 | } |
1481 | ||
1482 | /* | |
1483 | * Functions to lock and unlock any filesystem running on the | |
1484 | * device. | |
1485 | */ | |
2ca3310e | 1486 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1487 | { |
e39e2e95 | 1488 | int r; |
1da177e4 LT |
1489 | |
1490 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1491 | |
e39e2e95 | 1492 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1493 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1494 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1495 | md->frozen_sb = NULL; |
1496 | return r; | |
dfbe03f6 AK |
1497 | } |
1498 | ||
aa8d7c2f AK |
1499 | set_bit(DMF_FROZEN, &md->flags); |
1500 | ||
1da177e4 | 1501 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1502 | * to go away while it is locked. |
1da177e4 LT |
1503 | */ |
1504 | return 0; | |
1505 | } | |
1506 | ||
2ca3310e | 1507 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1508 | { |
aa8d7c2f AK |
1509 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1510 | return; | |
1511 | ||
e39e2e95 | 1512 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1513 | md->frozen_sb = NULL; |
aa8d7c2f | 1514 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1515 | } |
1516 | ||
1517 | /* | |
1518 | * We need to be able to change a mapping table under a mounted | |
1519 | * filesystem. For example we might want to move some data in | |
1520 | * the background. Before the table can be swapped with | |
1521 | * dm_bind_table, dm_suspend must be called to flush any in | |
1522 | * flight bios and ensure that any further io gets deferred. | |
1523 | */ | |
a3d77d35 | 1524 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1525 | { |
2ca3310e | 1526 | struct dm_table *map = NULL; |
46125c1c | 1527 | int r = 0; |
a3d77d35 | 1528 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1529 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1530 | |
e61290a4 | 1531 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1532 | |
73d410c0 MB |
1533 | if (dm_suspended(md)) { |
1534 | r = -EINVAL; | |
d287483d | 1535 | goto out_unlock; |
73d410c0 | 1536 | } |
1da177e4 LT |
1537 | |
1538 | map = dm_get_table(md); | |
1da177e4 | 1539 | |
2e93ccc1 KU |
1540 | /* |
1541 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1542 | * This flag is cleared before dm_suspend returns. | |
1543 | */ | |
1544 | if (noflush) | |
1545 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1546 | ||
cf222b37 AK |
1547 | /* This does not get reverted if there's an error later. */ |
1548 | dm_table_presuspend_targets(map); | |
1549 | ||
bfa152fa JN |
1550 | /* bdget() can stall if the pending I/Os are not flushed */ |
1551 | if (!noflush) { | |
1552 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1553 | if (!md->suspended_bdev) { | |
1554 | DMWARN("bdget failed in dm_suspend"); | |
1555 | r = -ENOMEM; | |
f431d966 | 1556 | goto out; |
bfa152fa | 1557 | } |
e39e2e95 | 1558 | |
6d6f10df MB |
1559 | /* |
1560 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1561 | * because lock_fs() needs to flush I/Os. | |
1562 | */ | |
1563 | if (do_lockfs) { | |
1564 | r = lock_fs(md); | |
1565 | if (r) | |
1566 | goto out; | |
1567 | } | |
aa8d7c2f | 1568 | } |
1da177e4 LT |
1569 | |
1570 | /* | |
354e0071 | 1571 | * First we set the BLOCK_IO flag so no more ios will be mapped. |
1da177e4 | 1572 | */ |
2ca3310e AK |
1573 | down_write(&md->io_lock); |
1574 | set_bit(DMF_BLOCK_IO, &md->flags); | |
1da177e4 | 1575 | |
2ca3310e | 1576 | up_write(&md->io_lock); |
1da177e4 | 1577 | |
1da177e4 | 1578 | /* |
46125c1c | 1579 | * Wait for the already-mapped ios to complete. |
1da177e4 | 1580 | */ |
401600df | 1581 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 1582 | |
2ca3310e | 1583 | down_write(&md->io_lock); |
1da177e4 | 1584 | |
6d6f10df | 1585 | if (noflush) |
022c2611 | 1586 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 1587 | up_write(&md->io_lock); |
2e93ccc1 | 1588 | |
1da177e4 | 1589 | /* were we interrupted ? */ |
46125c1c | 1590 | if (r < 0) { |
9a1fb464 | 1591 | dm_queue_flush(md); |
73d410c0 | 1592 | |
2ca3310e | 1593 | unlock_fs(md); |
2e93ccc1 | 1594 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1595 | } |
1da177e4 | 1596 | |
cf222b37 | 1597 | dm_table_postsuspend_targets(map); |
1da177e4 | 1598 | |
2ca3310e | 1599 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1600 | |
2ca3310e | 1601 | out: |
e39e2e95 AK |
1602 | if (r && md->suspended_bdev) { |
1603 | bdput(md->suspended_bdev); | |
1604 | md->suspended_bdev = NULL; | |
1605 | } | |
1606 | ||
2ca3310e | 1607 | dm_table_put(map); |
d287483d AK |
1608 | |
1609 | out_unlock: | |
e61290a4 | 1610 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1611 | return r; |
1da177e4 LT |
1612 | } |
1613 | ||
1614 | int dm_resume(struct mapped_device *md) | |
1615 | { | |
cf222b37 | 1616 | int r = -EINVAL; |
cf222b37 | 1617 | struct dm_table *map = NULL; |
1da177e4 | 1618 | |
e61290a4 | 1619 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1620 | if (!dm_suspended(md)) |
cf222b37 | 1621 | goto out; |
cf222b37 AK |
1622 | |
1623 | map = dm_get_table(md); | |
2ca3310e | 1624 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1625 | goto out; |
1da177e4 | 1626 | |
8757b776 MB |
1627 | r = dm_table_resume_targets(map); |
1628 | if (r) | |
1629 | goto out; | |
2ca3310e | 1630 | |
9a1fb464 | 1631 | dm_queue_flush(md); |
2ca3310e AK |
1632 | |
1633 | unlock_fs(md); | |
1634 | ||
bfa152fa JN |
1635 | if (md->suspended_bdev) { |
1636 | bdput(md->suspended_bdev); | |
1637 | md->suspended_bdev = NULL; | |
1638 | } | |
e39e2e95 | 1639 | |
2ca3310e AK |
1640 | clear_bit(DMF_SUSPENDED, &md->flags); |
1641 | ||
1da177e4 | 1642 | dm_table_unplug_all(map); |
1da177e4 | 1643 | |
69267a30 | 1644 | dm_kobject_uevent(md); |
8560ed6f | 1645 | |
cf222b37 | 1646 | r = 0; |
2ca3310e | 1647 | |
cf222b37 AK |
1648 | out: |
1649 | dm_table_put(map); | |
e61290a4 | 1650 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1651 | |
cf222b37 | 1652 | return r; |
1da177e4 LT |
1653 | } |
1654 | ||
1655 | /*----------------------------------------------------------------- | |
1656 | * Event notification. | |
1657 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1658 | void dm_kobject_uevent(struct mapped_device *md) |
1659 | { | |
ed9e1982 | 1660 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1661 | } |
1662 | ||
7a8c3d3b MA |
1663 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1664 | { | |
1665 | return atomic_add_return(1, &md->uevent_seq); | |
1666 | } | |
1667 | ||
1da177e4 LT |
1668 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1669 | { | |
1670 | return atomic_read(&md->event_nr); | |
1671 | } | |
1672 | ||
1673 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1674 | { | |
1675 | return wait_event_interruptible(md->eventq, | |
1676 | (event_nr != atomic_read(&md->event_nr))); | |
1677 | } | |
1678 | ||
7a8c3d3b MA |
1679 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1680 | { | |
1681 | unsigned long flags; | |
1682 | ||
1683 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1684 | list_add(elist, &md->uevent_list); | |
1685 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1686 | } | |
1687 | ||
1da177e4 LT |
1688 | /* |
1689 | * The gendisk is only valid as long as you have a reference | |
1690 | * count on 'md'. | |
1691 | */ | |
1692 | struct gendisk *dm_disk(struct mapped_device *md) | |
1693 | { | |
1694 | return md->disk; | |
1695 | } | |
1696 | ||
784aae73 MB |
1697 | struct kobject *dm_kobject(struct mapped_device *md) |
1698 | { | |
1699 | return &md->kobj; | |
1700 | } | |
1701 | ||
1702 | /* | |
1703 | * struct mapped_device should not be exported outside of dm.c | |
1704 | * so use this check to verify that kobj is part of md structure | |
1705 | */ | |
1706 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1707 | { | |
1708 | struct mapped_device *md; | |
1709 | ||
1710 | md = container_of(kobj, struct mapped_device, kobj); | |
1711 | if (&md->kobj != kobj) | |
1712 | return NULL; | |
1713 | ||
1714 | dm_get(md); | |
1715 | return md; | |
1716 | } | |
1717 | ||
1da177e4 LT |
1718 | int dm_suspended(struct mapped_device *md) |
1719 | { | |
1720 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1721 | } | |
1722 | ||
2e93ccc1 KU |
1723 | int dm_noflush_suspending(struct dm_target *ti) |
1724 | { | |
1725 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1726 | int r = __noflush_suspending(md); | |
1727 | ||
1728 | dm_put(md); | |
1729 | ||
1730 | return r; | |
1731 | } | |
1732 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1733 | ||
1da177e4 LT |
1734 | static struct block_device_operations dm_blk_dops = { |
1735 | .open = dm_blk_open, | |
1736 | .release = dm_blk_close, | |
aa129a22 | 1737 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1738 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1739 | .owner = THIS_MODULE |
1740 | }; | |
1741 | ||
1742 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1743 | ||
1744 | /* | |
1745 | * module hooks | |
1746 | */ | |
1747 | module_init(dm_init); | |
1748 | module_exit(dm_exit); | |
1749 | ||
1750 | module_param(major, uint, 0); | |
1751 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1752 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1753 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1754 | MODULE_LICENSE("GPL"); |