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 | */ | |
1eb787ec | 92 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
1da177e4 | 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 |
1eb787ec | 98 | #define DMF_QUEUE_IO_TO_THREAD 6 |
1da177e4 | 99 | |
304f3f6a MB |
100 | /* |
101 | * Work processed by per-device workqueue. | |
102 | */ | |
1da177e4 | 103 | struct mapped_device { |
2ca3310e | 104 | struct rw_semaphore io_lock; |
e61290a4 | 105 | struct mutex suspend_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; |
022c2611 | 125 | spinlock_t deferred_lock; |
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 | */ | |
92c63902 | 439 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 440 | { |
2ca3310e | 441 | down_write(&md->io_lock); |
1da177e4 | 442 | |
022c2611 | 443 | spin_lock_irq(&md->deferred_lock); |
1da177e4 | 444 | bio_list_add(&md->deferred, bio); |
022c2611 | 445 | spin_unlock_irq(&md->deferred_lock); |
1da177e4 | 446 | |
92c63902 MP |
447 | if (!test_and_set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) |
448 | queue_work(md->wq, &md->work); | |
449 | ||
2ca3310e | 450 | up_write(&md->io_lock); |
1da177e4 LT |
451 | } |
452 | ||
453 | /* | |
454 | * Everyone (including functions in this file), should use this | |
455 | * function to access the md->map field, and make sure they call | |
456 | * dm_table_put() when finished. | |
457 | */ | |
458 | struct dm_table *dm_get_table(struct mapped_device *md) | |
459 | { | |
460 | struct dm_table *t; | |
461 | ||
462 | read_lock(&md->map_lock); | |
463 | t = md->map; | |
464 | if (t) | |
465 | dm_table_get(t); | |
466 | read_unlock(&md->map_lock); | |
467 | ||
468 | return t; | |
469 | } | |
470 | ||
3ac51e74 DW |
471 | /* |
472 | * Get the geometry associated with a dm device | |
473 | */ | |
474 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
475 | { | |
476 | *geo = md->geometry; | |
477 | ||
478 | return 0; | |
479 | } | |
480 | ||
481 | /* | |
482 | * Set the geometry of a device. | |
483 | */ | |
484 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
485 | { | |
486 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
487 | ||
488 | if (geo->start > sz) { | |
489 | DMWARN("Start sector is beyond the geometry limits."); | |
490 | return -EINVAL; | |
491 | } | |
492 | ||
493 | md->geometry = *geo; | |
494 | ||
495 | return 0; | |
496 | } | |
497 | ||
1da177e4 LT |
498 | /*----------------------------------------------------------------- |
499 | * CRUD START: | |
500 | * A more elegant soln is in the works that uses the queue | |
501 | * merge fn, unfortunately there are a couple of changes to | |
502 | * the block layer that I want to make for this. So in the | |
503 | * interests of getting something for people to use I give | |
504 | * you this clearly demarcated crap. | |
505 | *---------------------------------------------------------------*/ | |
506 | ||
2e93ccc1 KU |
507 | static int __noflush_suspending(struct mapped_device *md) |
508 | { | |
509 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
510 | } | |
511 | ||
1da177e4 LT |
512 | /* |
513 | * Decrements the number of outstanding ios that a bio has been | |
514 | * cloned into, completing the original io if necc. | |
515 | */ | |
858119e1 | 516 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 517 | { |
2e93ccc1 | 518 | unsigned long flags; |
b35f8caa MB |
519 | int io_error; |
520 | struct bio *bio; | |
521 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
522 | |
523 | /* Push-back supersedes any I/O errors */ | |
b35f8caa | 524 | if (error && !(io->error > 0 && __noflush_suspending(md))) |
1da177e4 LT |
525 | io->error = error; |
526 | ||
527 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
528 | if (io->error == DM_ENDIO_REQUEUE) { |
529 | /* | |
530 | * Target requested pushing back the I/O. | |
2e93ccc1 | 531 | */ |
022c2611 | 532 | spin_lock_irqsave(&md->deferred_lock, flags); |
b35f8caa | 533 | if (__noflush_suspending(md)) |
022c2611 | 534 | bio_list_add(&md->deferred, io->bio); |
2e93ccc1 KU |
535 | else |
536 | /* noflush suspend was interrupted. */ | |
537 | io->error = -EIO; | |
022c2611 | 538 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
539 | } |
540 | ||
d221d2e7 | 541 | end_io_acct(io); |
1da177e4 | 542 | |
b35f8caa MB |
543 | io_error = io->error; |
544 | bio = io->bio; | |
2e93ccc1 | 545 | |
b35f8caa MB |
546 | free_io(md, io); |
547 | ||
548 | if (io_error != DM_ENDIO_REQUEUE) { | |
549 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 550 | |
b35f8caa MB |
551 | bio_endio(bio, io_error); |
552 | } | |
1da177e4 LT |
553 | } |
554 | } | |
555 | ||
6712ecf8 | 556 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
557 | { |
558 | int r = 0; | |
028867ac | 559 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 560 | struct dm_io *io = tio->io; |
9faf400f | 561 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
562 | dm_endio_fn endio = tio->ti->type->end_io; |
563 | ||
1da177e4 LT |
564 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
565 | error = -EIO; | |
566 | ||
567 | if (endio) { | |
568 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
569 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
570 | /* | |
571 | * error and requeue request are handled | |
572 | * in dec_pending(). | |
573 | */ | |
1da177e4 | 574 | error = r; |
45cbcd79 KU |
575 | else if (r == DM_ENDIO_INCOMPLETE) |
576 | /* The target will handle the io */ | |
6712ecf8 | 577 | return; |
45cbcd79 KU |
578 | else if (r) { |
579 | DMWARN("unimplemented target endio return value: %d", r); | |
580 | BUG(); | |
581 | } | |
1da177e4 LT |
582 | } |
583 | ||
9faf400f SB |
584 | /* |
585 | * Store md for cleanup instead of tio which is about to get freed. | |
586 | */ | |
587 | bio->bi_private = md->bs; | |
588 | ||
9faf400f | 589 | free_tio(md, tio); |
b35f8caa MB |
590 | bio_put(bio); |
591 | dec_pending(io, error); | |
1da177e4 LT |
592 | } |
593 | ||
594 | static sector_t max_io_len(struct mapped_device *md, | |
595 | sector_t sector, struct dm_target *ti) | |
596 | { | |
597 | sector_t offset = sector - ti->begin; | |
598 | sector_t len = ti->len - offset; | |
599 | ||
600 | /* | |
601 | * Does the target need to split even further ? | |
602 | */ | |
603 | if (ti->split_io) { | |
604 | sector_t boundary; | |
605 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
606 | - offset; | |
607 | if (len > boundary) | |
608 | len = boundary; | |
609 | } | |
610 | ||
611 | return len; | |
612 | } | |
613 | ||
614 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 615 | struct dm_target_io *tio) |
1da177e4 LT |
616 | { |
617 | int r; | |
2056a782 | 618 | sector_t sector; |
9faf400f | 619 | struct mapped_device *md; |
1da177e4 LT |
620 | |
621 | /* | |
622 | * Sanity checks. | |
623 | */ | |
624 | BUG_ON(!clone->bi_size); | |
625 | ||
626 | clone->bi_end_io = clone_endio; | |
627 | clone->bi_private = tio; | |
628 | ||
629 | /* | |
630 | * Map the clone. If r == 0 we don't need to do | |
631 | * anything, the target has assumed ownership of | |
632 | * this io. | |
633 | */ | |
634 | atomic_inc(&tio->io->io_count); | |
2056a782 | 635 | sector = clone->bi_sector; |
1da177e4 | 636 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 637 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 638 | /* the bio has been remapped so dispatch it */ |
2056a782 | 639 | |
5f3ea37c | 640 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
641 | tio->io->bio->bi_bdev->bd_dev, |
642 | clone->bi_sector, sector); | |
2056a782 | 643 | |
1da177e4 | 644 | generic_make_request(clone); |
2e93ccc1 KU |
645 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
646 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
647 | md = tio->io->md; |
648 | dec_pending(tio->io, r); | |
649 | /* | |
650 | * Store bio_set for cleanup. | |
651 | */ | |
652 | clone->bi_private = md->bs; | |
1da177e4 | 653 | bio_put(clone); |
9faf400f | 654 | free_tio(md, tio); |
45cbcd79 KU |
655 | } else if (r) { |
656 | DMWARN("unimplemented target map return value: %d", r); | |
657 | BUG(); | |
1da177e4 LT |
658 | } |
659 | } | |
660 | ||
661 | struct clone_info { | |
662 | struct mapped_device *md; | |
663 | struct dm_table *map; | |
664 | struct bio *bio; | |
665 | struct dm_io *io; | |
666 | sector_t sector; | |
667 | sector_t sector_count; | |
668 | unsigned short idx; | |
669 | }; | |
670 | ||
3676347a PO |
671 | static void dm_bio_destructor(struct bio *bio) |
672 | { | |
9faf400f SB |
673 | struct bio_set *bs = bio->bi_private; |
674 | ||
675 | bio_free(bio, bs); | |
3676347a PO |
676 | } |
677 | ||
1da177e4 LT |
678 | /* |
679 | * Creates a little bio that is just does part of a bvec. | |
680 | */ | |
681 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
682 | unsigned short idx, unsigned int offset, | |
9faf400f | 683 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
684 | { |
685 | struct bio *clone; | |
686 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
687 | ||
9faf400f | 688 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 689 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
690 | *clone->bi_io_vec = *bv; |
691 | ||
692 | clone->bi_sector = sector; | |
693 | clone->bi_bdev = bio->bi_bdev; | |
694 | clone->bi_rw = bio->bi_rw; | |
695 | clone->bi_vcnt = 1; | |
696 | clone->bi_size = to_bytes(len); | |
697 | clone->bi_io_vec->bv_offset = offset; | |
698 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 699 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 700 | |
9c47008d MP |
701 | if (bio_integrity(bio)) { |
702 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
703 | bio_integrity_trim(clone, | |
704 | bio_sector_offset(bio, idx, offset), len); | |
705 | } | |
706 | ||
1da177e4 LT |
707 | return clone; |
708 | } | |
709 | ||
710 | /* | |
711 | * Creates a bio that consists of range of complete bvecs. | |
712 | */ | |
713 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
714 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 715 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
716 | { |
717 | struct bio *clone; | |
718 | ||
9faf400f SB |
719 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
720 | __bio_clone(clone, bio); | |
721 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
722 | clone->bi_sector = sector; |
723 | clone->bi_idx = idx; | |
724 | clone->bi_vcnt = idx + bv_count; | |
725 | clone->bi_size = to_bytes(len); | |
726 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
727 | ||
9c47008d MP |
728 | if (bio_integrity(bio)) { |
729 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
730 | ||
731 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
732 | bio_integrity_trim(clone, | |
733 | bio_sector_offset(bio, idx, 0), len); | |
734 | } | |
735 | ||
1da177e4 LT |
736 | return clone; |
737 | } | |
738 | ||
512875bd | 739 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
740 | { |
741 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
742 | struct dm_target *ti; |
743 | sector_t len = 0, max; | |
028867ac | 744 | struct dm_target_io *tio; |
1da177e4 | 745 | |
512875bd JN |
746 | ti = dm_table_find_target(ci->map, ci->sector); |
747 | if (!dm_target_is_valid(ti)) | |
748 | return -EIO; | |
749 | ||
750 | max = max_io_len(ci->md, ci->sector, ti); | |
751 | ||
1da177e4 LT |
752 | /* |
753 | * Allocate a target io object. | |
754 | */ | |
755 | tio = alloc_tio(ci->md); | |
756 | tio->io = ci->io; | |
757 | tio->ti = ti; | |
758 | memset(&tio->info, 0, sizeof(tio->info)); | |
759 | ||
760 | if (ci->sector_count <= max) { | |
761 | /* | |
762 | * Optimise for the simple case where we can do all of | |
763 | * the remaining io with a single clone. | |
764 | */ | |
765 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
766 | bio->bi_vcnt - ci->idx, ci->sector_count, |
767 | ci->md->bs); | |
1da177e4 LT |
768 | __map_bio(ti, clone, tio); |
769 | ci->sector_count = 0; | |
770 | ||
771 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
772 | /* | |
773 | * There are some bvecs that don't span targets. | |
774 | * Do as many of these as possible. | |
775 | */ | |
776 | int i; | |
777 | sector_t remaining = max; | |
778 | sector_t bv_len; | |
779 | ||
780 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
781 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
782 | ||
783 | if (bv_len > remaining) | |
784 | break; | |
785 | ||
786 | remaining -= bv_len; | |
787 | len += bv_len; | |
788 | } | |
789 | ||
9faf400f SB |
790 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
791 | ci->md->bs); | |
1da177e4 LT |
792 | __map_bio(ti, clone, tio); |
793 | ||
794 | ci->sector += len; | |
795 | ci->sector_count -= len; | |
796 | ci->idx = i; | |
797 | ||
798 | } else { | |
799 | /* | |
d2044a94 | 800 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
801 | */ |
802 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
803 | sector_t remaining = to_sector(bv->bv_len); |
804 | unsigned int offset = 0; | |
1da177e4 | 805 | |
d2044a94 AK |
806 | do { |
807 | if (offset) { | |
808 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
809 | if (!dm_target_is_valid(ti)) |
810 | return -EIO; | |
811 | ||
d2044a94 | 812 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 813 | |
d2044a94 AK |
814 | tio = alloc_tio(ci->md); |
815 | tio->io = ci->io; | |
816 | tio->ti = ti; | |
817 | memset(&tio->info, 0, sizeof(tio->info)); | |
818 | } | |
819 | ||
820 | len = min(remaining, max); | |
821 | ||
822 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
823 | bv->bv_offset + offset, len, |
824 | ci->md->bs); | |
d2044a94 AK |
825 | |
826 | __map_bio(ti, clone, tio); | |
827 | ||
828 | ci->sector += len; | |
829 | ci->sector_count -= len; | |
830 | offset += to_bytes(len); | |
831 | } while (remaining -= len); | |
1da177e4 | 832 | |
1da177e4 LT |
833 | ci->idx++; |
834 | } | |
512875bd JN |
835 | |
836 | return 0; | |
1da177e4 LT |
837 | } |
838 | ||
839 | /* | |
8a53c28d | 840 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 841 | */ |
f0b9a450 | 842 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
843 | { |
844 | struct clone_info ci; | |
512875bd | 845 | int error = 0; |
1da177e4 LT |
846 | |
847 | ci.map = dm_get_table(md); | |
f0b9a450 MP |
848 | if (unlikely(!ci.map)) { |
849 | bio_io_error(bio); | |
850 | return; | |
851 | } | |
692d0eb9 | 852 | |
1da177e4 LT |
853 | ci.md = md; |
854 | ci.bio = bio; | |
855 | ci.io = alloc_io(md); | |
856 | ci.io->error = 0; | |
857 | atomic_set(&ci.io->io_count, 1); | |
858 | ci.io->bio = bio; | |
859 | ci.io->md = md; | |
860 | ci.sector = bio->bi_sector; | |
861 | ci.sector_count = bio_sectors(bio); | |
862 | ci.idx = bio->bi_idx; | |
863 | ||
3eaf840e | 864 | start_io_acct(ci.io); |
512875bd JN |
865 | while (ci.sector_count && !error) |
866 | error = __clone_and_map(&ci); | |
1da177e4 LT |
867 | |
868 | /* drop the extra reference count */ | |
512875bd | 869 | dec_pending(ci.io, error); |
1da177e4 LT |
870 | dm_table_put(ci.map); |
871 | } | |
872 | /*----------------------------------------------------------------- | |
873 | * CRUD END | |
874 | *---------------------------------------------------------------*/ | |
875 | ||
f6fccb12 MB |
876 | static int dm_merge_bvec(struct request_queue *q, |
877 | struct bvec_merge_data *bvm, | |
878 | struct bio_vec *biovec) | |
879 | { | |
880 | struct mapped_device *md = q->queuedata; | |
881 | struct dm_table *map = dm_get_table(md); | |
882 | struct dm_target *ti; | |
883 | sector_t max_sectors; | |
5037108a | 884 | int max_size = 0; |
f6fccb12 MB |
885 | |
886 | if (unlikely(!map)) | |
5037108a | 887 | goto out; |
f6fccb12 MB |
888 | |
889 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
890 | if (!dm_target_is_valid(ti)) |
891 | goto out_table; | |
f6fccb12 MB |
892 | |
893 | /* | |
894 | * Find maximum amount of I/O that won't need splitting | |
895 | */ | |
896 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
897 | (sector_t) BIO_MAX_SECTORS); | |
898 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
899 | if (max_size < 0) | |
900 | max_size = 0; | |
901 | ||
902 | /* | |
903 | * merge_bvec_fn() returns number of bytes | |
904 | * it can accept at this offset | |
905 | * max is precomputed maximal io size | |
906 | */ | |
907 | if (max_size && ti->type->merge) | |
908 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
909 | ||
b01cd5ac | 910 | out_table: |
5037108a MP |
911 | dm_table_put(map); |
912 | ||
913 | out: | |
f6fccb12 MB |
914 | /* |
915 | * Always allow an entire first page | |
916 | */ | |
917 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
918 | max_size = biovec->bv_len; | |
919 | ||
f6fccb12 MB |
920 | return max_size; |
921 | } | |
922 | ||
1da177e4 LT |
923 | /* |
924 | * The request function that just remaps the bio built up by | |
925 | * dm_merge_bvec. | |
926 | */ | |
165125e1 | 927 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 928 | { |
12f03a49 | 929 | int rw = bio_data_dir(bio); |
1da177e4 | 930 | struct mapped_device *md = q->queuedata; |
c9959059 | 931 | int cpu; |
1da177e4 | 932 | |
692d0eb9 MP |
933 | /* |
934 | * There is no use in forwarding any barrier request since we can't | |
935 | * guarantee it is (or can be) handled by the targets correctly. | |
936 | */ | |
937 | if (unlikely(bio_barrier(bio))) { | |
938 | bio_endio(bio, -EOPNOTSUPP); | |
939 | return 0; | |
940 | } | |
941 | ||
2ca3310e | 942 | down_read(&md->io_lock); |
1da177e4 | 943 | |
074a7aca TH |
944 | cpu = part_stat_lock(); |
945 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
946 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
947 | part_stat_unlock(); | |
12f03a49 | 948 | |
1da177e4 | 949 | /* |
1eb787ec AK |
950 | * If we're suspended or the thread is processing barriers |
951 | * we have to queue this io for later. | |
1da177e4 | 952 | */ |
92c63902 | 953 | if (unlikely(test_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags))) { |
2ca3310e | 954 | up_read(&md->io_lock); |
1da177e4 | 955 | |
54d9a1b4 AK |
956 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) && |
957 | bio_rw(bio) == READA) { | |
958 | bio_io_error(bio); | |
959 | return 0; | |
960 | } | |
1da177e4 | 961 | |
92c63902 | 962 | queue_io(md, bio); |
1da177e4 | 963 | |
92c63902 | 964 | return 0; |
1da177e4 LT |
965 | } |
966 | ||
f0b9a450 | 967 | __split_and_process_bio(md, bio); |
2ca3310e | 968 | up_read(&md->io_lock); |
f0b9a450 | 969 | return 0; |
1da177e4 LT |
970 | } |
971 | ||
165125e1 | 972 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
973 | { |
974 | struct mapped_device *md = q->queuedata; | |
975 | struct dm_table *map = dm_get_table(md); | |
976 | ||
977 | if (map) { | |
978 | dm_table_unplug_all(map); | |
979 | dm_table_put(map); | |
980 | } | |
981 | } | |
982 | ||
983 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
984 | { | |
8a57dfc6 CS |
985 | int r = bdi_bits; |
986 | struct mapped_device *md = congested_data; | |
987 | struct dm_table *map; | |
1da177e4 | 988 | |
1eb787ec | 989 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
8a57dfc6 CS |
990 | map = dm_get_table(md); |
991 | if (map) { | |
992 | r = dm_table_any_congested(map, bdi_bits); | |
993 | dm_table_put(map); | |
994 | } | |
995 | } | |
996 | ||
1da177e4 LT |
997 | return r; |
998 | } | |
999 | ||
1000 | /*----------------------------------------------------------------- | |
1001 | * An IDR is used to keep track of allocated minor numbers. | |
1002 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1003 | static DEFINE_IDR(_minor_idr); |
1004 | ||
2b06cfff | 1005 | static void free_minor(int minor) |
1da177e4 | 1006 | { |
f32c10b0 | 1007 | spin_lock(&_minor_lock); |
1da177e4 | 1008 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1009 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1010 | } |
1011 | ||
1012 | /* | |
1013 | * See if the device with a specific minor # is free. | |
1014 | */ | |
cf13ab8e | 1015 | static int specific_minor(int minor) |
1da177e4 LT |
1016 | { |
1017 | int r, m; | |
1018 | ||
1019 | if (minor >= (1 << MINORBITS)) | |
1020 | return -EINVAL; | |
1021 | ||
62f75c2f JM |
1022 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1023 | if (!r) | |
1024 | return -ENOMEM; | |
1025 | ||
f32c10b0 | 1026 | spin_lock(&_minor_lock); |
1da177e4 LT |
1027 | |
1028 | if (idr_find(&_minor_idr, minor)) { | |
1029 | r = -EBUSY; | |
1030 | goto out; | |
1031 | } | |
1032 | ||
ba61fdd1 | 1033 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1034 | if (r) |
1da177e4 | 1035 | goto out; |
1da177e4 LT |
1036 | |
1037 | if (m != minor) { | |
1038 | idr_remove(&_minor_idr, m); | |
1039 | r = -EBUSY; | |
1040 | goto out; | |
1041 | } | |
1042 | ||
1043 | out: | |
f32c10b0 | 1044 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1045 | return r; |
1046 | } | |
1047 | ||
cf13ab8e | 1048 | static int next_free_minor(int *minor) |
1da177e4 | 1049 | { |
2b06cfff | 1050 | int r, m; |
1da177e4 | 1051 | |
1da177e4 | 1052 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1053 | if (!r) |
1054 | return -ENOMEM; | |
1055 | ||
f32c10b0 | 1056 | spin_lock(&_minor_lock); |
1da177e4 | 1057 | |
ba61fdd1 | 1058 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1059 | if (r) |
1da177e4 | 1060 | goto out; |
1da177e4 LT |
1061 | |
1062 | if (m >= (1 << MINORBITS)) { | |
1063 | idr_remove(&_minor_idr, m); | |
1064 | r = -ENOSPC; | |
1065 | goto out; | |
1066 | } | |
1067 | ||
1068 | *minor = m; | |
1069 | ||
1070 | out: | |
f32c10b0 | 1071 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1072 | return r; |
1073 | } | |
1074 | ||
1075 | static struct block_device_operations dm_blk_dops; | |
1076 | ||
53d5914f MP |
1077 | static void dm_wq_work(struct work_struct *work); |
1078 | ||
1da177e4 LT |
1079 | /* |
1080 | * Allocate and initialise a blank device with a given minor. | |
1081 | */ | |
2b06cfff | 1082 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1083 | { |
1084 | int r; | |
cf13ab8e | 1085 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1086 | void *old_md; |
1da177e4 LT |
1087 | |
1088 | if (!md) { | |
1089 | DMWARN("unable to allocate device, out of memory."); | |
1090 | return NULL; | |
1091 | } | |
1092 | ||
10da4f79 | 1093 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1094 | goto bad_module_get; |
10da4f79 | 1095 | |
1da177e4 | 1096 | /* get a minor number for the dev */ |
2b06cfff | 1097 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1098 | r = next_free_minor(&minor); |
2b06cfff | 1099 | else |
cf13ab8e | 1100 | r = specific_minor(minor); |
1da177e4 | 1101 | if (r < 0) |
6ed7ade8 | 1102 | goto bad_minor; |
1da177e4 | 1103 | |
2ca3310e | 1104 | init_rwsem(&md->io_lock); |
e61290a4 | 1105 | mutex_init(&md->suspend_lock); |
022c2611 | 1106 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1107 | rwlock_init(&md->map_lock); |
1108 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1109 | atomic_set(&md->open_count, 0); |
1da177e4 | 1110 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1111 | atomic_set(&md->uevent_seq, 0); |
1112 | INIT_LIST_HEAD(&md->uevent_list); | |
1113 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1114 | |
1115 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1116 | if (!md->queue) | |
6ed7ade8 | 1117 | goto bad_queue; |
1da177e4 LT |
1118 | |
1119 | md->queue->queuedata = md; | |
1120 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1121 | md->queue->backing_dev_info.congested_data = md; | |
1122 | blk_queue_make_request(md->queue, dm_request); | |
99360b4c | 1123 | blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN, NULL); |
daef265f | 1124 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1125 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1126 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1127 | |
93d2341c | 1128 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1129 | if (!md->io_pool) |
6ed7ade8 | 1130 | goto bad_io_pool; |
1da177e4 | 1131 | |
93d2341c | 1132 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1133 | if (!md->tio_pool) |
6ed7ade8 | 1134 | goto bad_tio_pool; |
1da177e4 | 1135 | |
bb799ca0 | 1136 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1137 | if (!md->bs) |
1138 | goto bad_no_bioset; | |
1139 | ||
1da177e4 LT |
1140 | md->disk = alloc_disk(1); |
1141 | if (!md->disk) | |
6ed7ade8 | 1142 | goto bad_disk; |
1da177e4 | 1143 | |
f0b04115 JM |
1144 | atomic_set(&md->pending, 0); |
1145 | init_waitqueue_head(&md->wait); | |
53d5914f | 1146 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1147 | init_waitqueue_head(&md->eventq); |
1148 | ||
1da177e4 LT |
1149 | md->disk->major = _major; |
1150 | md->disk->first_minor = minor; | |
1151 | md->disk->fops = &dm_blk_dops; | |
1152 | md->disk->queue = md->queue; | |
1153 | md->disk->private_data = md; | |
1154 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1155 | add_disk(md->disk); | |
7e51f257 | 1156 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1157 | |
304f3f6a MB |
1158 | md->wq = create_singlethread_workqueue("kdmflush"); |
1159 | if (!md->wq) | |
1160 | goto bad_thread; | |
1161 | ||
ba61fdd1 | 1162 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1163 | spin_lock(&_minor_lock); |
ba61fdd1 | 1164 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1165 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1166 | |
1167 | BUG_ON(old_md != MINOR_ALLOCED); | |
1168 | ||
1da177e4 LT |
1169 | return md; |
1170 | ||
304f3f6a MB |
1171 | bad_thread: |
1172 | put_disk(md->disk); | |
6ed7ade8 | 1173 | bad_disk: |
9faf400f | 1174 | bioset_free(md->bs); |
6ed7ade8 | 1175 | bad_no_bioset: |
1da177e4 | 1176 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1177 | bad_tio_pool: |
1da177e4 | 1178 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1179 | bad_io_pool: |
1312f40e | 1180 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1181 | bad_queue: |
1da177e4 | 1182 | free_minor(minor); |
6ed7ade8 | 1183 | bad_minor: |
10da4f79 | 1184 | module_put(THIS_MODULE); |
6ed7ade8 | 1185 | bad_module_get: |
1da177e4 LT |
1186 | kfree(md); |
1187 | return NULL; | |
1188 | } | |
1189 | ||
ae9da83f JN |
1190 | static void unlock_fs(struct mapped_device *md); |
1191 | ||
1da177e4 LT |
1192 | static void free_dev(struct mapped_device *md) |
1193 | { | |
f331c029 | 1194 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1195 | |
d9dde59b | 1196 | if (md->suspended_bdev) { |
ae9da83f | 1197 | unlock_fs(md); |
d9dde59b JN |
1198 | bdput(md->suspended_bdev); |
1199 | } | |
304f3f6a | 1200 | destroy_workqueue(md->wq); |
1da177e4 LT |
1201 | mempool_destroy(md->tio_pool); |
1202 | mempool_destroy(md->io_pool); | |
9faf400f | 1203 | bioset_free(md->bs); |
9c47008d | 1204 | blk_integrity_unregister(md->disk); |
1da177e4 | 1205 | del_gendisk(md->disk); |
63d94e48 | 1206 | free_minor(minor); |
fba9f90e JM |
1207 | |
1208 | spin_lock(&_minor_lock); | |
1209 | md->disk->private_data = NULL; | |
1210 | spin_unlock(&_minor_lock); | |
1211 | ||
1da177e4 | 1212 | put_disk(md->disk); |
1312f40e | 1213 | blk_cleanup_queue(md->queue); |
10da4f79 | 1214 | module_put(THIS_MODULE); |
1da177e4 LT |
1215 | kfree(md); |
1216 | } | |
1217 | ||
1218 | /* | |
1219 | * Bind a table to the device. | |
1220 | */ | |
1221 | static void event_callback(void *context) | |
1222 | { | |
7a8c3d3b MA |
1223 | unsigned long flags; |
1224 | LIST_HEAD(uevents); | |
1da177e4 LT |
1225 | struct mapped_device *md = (struct mapped_device *) context; |
1226 | ||
7a8c3d3b MA |
1227 | spin_lock_irqsave(&md->uevent_lock, flags); |
1228 | list_splice_init(&md->uevent_list, &uevents); | |
1229 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1230 | ||
ed9e1982 | 1231 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1232 | |
1da177e4 LT |
1233 | atomic_inc(&md->event_nr); |
1234 | wake_up(&md->eventq); | |
1235 | } | |
1236 | ||
4e90188b | 1237 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1238 | { |
4e90188b | 1239 | set_capacity(md->disk, size); |
1da177e4 | 1240 | |
1b1dcc1b | 1241 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1242 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1243 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1244 | } |
1245 | ||
1246 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1247 | { | |
165125e1 | 1248 | struct request_queue *q = md->queue; |
1da177e4 LT |
1249 | sector_t size; |
1250 | ||
1251 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1252 | |
1253 | /* | |
1254 | * Wipe any geometry if the size of the table changed. | |
1255 | */ | |
1256 | if (size != get_capacity(md->disk)) | |
1257 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1258 | ||
bfa152fa JN |
1259 | if (md->suspended_bdev) |
1260 | __set_size(md, size); | |
d5816876 MP |
1261 | |
1262 | if (!size) { | |
1263 | dm_table_destroy(t); | |
1da177e4 | 1264 | return 0; |
d5816876 | 1265 | } |
1da177e4 | 1266 | |
2ca3310e AK |
1267 | dm_table_event_callback(t, event_callback, md); |
1268 | ||
1da177e4 LT |
1269 | write_lock(&md->map_lock); |
1270 | md->map = t; | |
2ca3310e | 1271 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1272 | write_unlock(&md->map_lock); |
1273 | ||
1da177e4 LT |
1274 | return 0; |
1275 | } | |
1276 | ||
1277 | static void __unbind(struct mapped_device *md) | |
1278 | { | |
1279 | struct dm_table *map = md->map; | |
1280 | ||
1281 | if (!map) | |
1282 | return; | |
1283 | ||
1284 | dm_table_event_callback(map, NULL, NULL); | |
1285 | write_lock(&md->map_lock); | |
1286 | md->map = NULL; | |
1287 | write_unlock(&md->map_lock); | |
d5816876 | 1288 | dm_table_destroy(map); |
1da177e4 LT |
1289 | } |
1290 | ||
1291 | /* | |
1292 | * Constructor for a new device. | |
1293 | */ | |
2b06cfff | 1294 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1295 | { |
1296 | struct mapped_device *md; | |
1297 | ||
2b06cfff | 1298 | md = alloc_dev(minor); |
1da177e4 LT |
1299 | if (!md) |
1300 | return -ENXIO; | |
1301 | ||
784aae73 MB |
1302 | dm_sysfs_init(md); |
1303 | ||
1da177e4 LT |
1304 | *result = md; |
1305 | return 0; | |
1306 | } | |
1307 | ||
637842cf | 1308 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1309 | { |
1310 | struct mapped_device *md; | |
1da177e4 LT |
1311 | unsigned minor = MINOR(dev); |
1312 | ||
1313 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1314 | return NULL; | |
1315 | ||
f32c10b0 | 1316 | spin_lock(&_minor_lock); |
1da177e4 LT |
1317 | |
1318 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1319 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1320 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1321 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1322 | md = NULL; |
fba9f90e JM |
1323 | goto out; |
1324 | } | |
1da177e4 | 1325 | |
fba9f90e | 1326 | out: |
f32c10b0 | 1327 | spin_unlock(&_minor_lock); |
1da177e4 | 1328 | |
637842cf DT |
1329 | return md; |
1330 | } | |
1331 | ||
d229a958 DT |
1332 | struct mapped_device *dm_get_md(dev_t dev) |
1333 | { | |
1334 | struct mapped_device *md = dm_find_md(dev); | |
1335 | ||
1336 | if (md) | |
1337 | dm_get(md); | |
1338 | ||
1339 | return md; | |
1340 | } | |
1341 | ||
9ade92a9 | 1342 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1343 | { |
9ade92a9 | 1344 | return md->interface_ptr; |
1da177e4 LT |
1345 | } |
1346 | ||
1347 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1348 | { | |
1349 | md->interface_ptr = ptr; | |
1350 | } | |
1351 | ||
1352 | void dm_get(struct mapped_device *md) | |
1353 | { | |
1354 | atomic_inc(&md->holders); | |
1355 | } | |
1356 | ||
72d94861 AK |
1357 | const char *dm_device_name(struct mapped_device *md) |
1358 | { | |
1359 | return md->name; | |
1360 | } | |
1361 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1362 | ||
1da177e4 LT |
1363 | void dm_put(struct mapped_device *md) |
1364 | { | |
1134e5ae | 1365 | struct dm_table *map; |
1da177e4 | 1366 | |
fba9f90e JM |
1367 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1368 | ||
f32c10b0 | 1369 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1370 | map = dm_get_table(md); |
f331c029 TH |
1371 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1372 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1373 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1374 | spin_unlock(&_minor_lock); |
cf222b37 | 1375 | if (!dm_suspended(md)) { |
1da177e4 LT |
1376 | dm_table_presuspend_targets(map); |
1377 | dm_table_postsuspend_targets(map); | |
1378 | } | |
784aae73 | 1379 | dm_sysfs_exit(md); |
1134e5ae | 1380 | dm_table_put(map); |
a1b51e98 | 1381 | __unbind(md); |
1da177e4 LT |
1382 | free_dev(md); |
1383 | } | |
1da177e4 | 1384 | } |
79eb885c | 1385 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1386 | |
401600df | 1387 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
1388 | { |
1389 | int r = 0; | |
b44ebeb0 MP |
1390 | DECLARE_WAITQUEUE(wait, current); |
1391 | ||
1392 | dm_unplug_all(md->queue); | |
1393 | ||
1394 | add_wait_queue(&md->wait, &wait); | |
46125c1c MB |
1395 | |
1396 | while (1) { | |
401600df | 1397 | set_current_state(interruptible); |
46125c1c MB |
1398 | |
1399 | smp_mb(); | |
1400 | if (!atomic_read(&md->pending)) | |
1401 | break; | |
1402 | ||
401600df MP |
1403 | if (interruptible == TASK_INTERRUPTIBLE && |
1404 | signal_pending(current)) { | |
46125c1c MB |
1405 | r = -EINTR; |
1406 | break; | |
1407 | } | |
1408 | ||
1409 | io_schedule(); | |
1410 | } | |
1411 | set_current_state(TASK_RUNNING); | |
1412 | ||
b44ebeb0 MP |
1413 | remove_wait_queue(&md->wait, &wait); |
1414 | ||
46125c1c MB |
1415 | return r; |
1416 | } | |
1417 | ||
1da177e4 LT |
1418 | /* |
1419 | * Process the deferred bios | |
1420 | */ | |
ef208587 | 1421 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 1422 | { |
ef208587 MP |
1423 | struct mapped_device *md = container_of(work, struct mapped_device, |
1424 | work); | |
6d6f10df | 1425 | struct bio *c; |
1da177e4 | 1426 | |
ef208587 MP |
1427 | down_write(&md->io_lock); |
1428 | ||
3b00b203 | 1429 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
1430 | spin_lock_irq(&md->deferred_lock); |
1431 | c = bio_list_pop(&md->deferred); | |
1432 | spin_unlock_irq(&md->deferred_lock); | |
1433 | ||
1434 | if (!c) { | |
1eb787ec | 1435 | clear_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); |
df12ee99 AK |
1436 | break; |
1437 | } | |
022c2611 | 1438 | |
3b00b203 MP |
1439 | up_write(&md->io_lock); |
1440 | ||
f0b9a450 | 1441 | __split_and_process_bio(md, c); |
3b00b203 MP |
1442 | |
1443 | down_write(&md->io_lock); | |
022c2611 | 1444 | } |
73d410c0 | 1445 | |
ef208587 | 1446 | up_write(&md->io_lock); |
1da177e4 LT |
1447 | } |
1448 | ||
9a1fb464 | 1449 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 1450 | { |
3b00b203 MP |
1451 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1452 | smp_mb__after_clear_bit(); | |
53d5914f | 1453 | queue_work(md->wq, &md->work); |
304f3f6a MB |
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 | /* | |
3b00b203 MP |
1571 | * Here we must make sure that no processes are submitting requests |
1572 | * to target drivers i.e. no one may be executing | |
1573 | * __split_and_process_bio. This is called from dm_request and | |
1574 | * dm_wq_work. | |
1575 | * | |
1576 | * To get all processes out of __split_and_process_bio in dm_request, | |
1577 | * we take the write lock. To prevent any process from reentering | |
1578 | * __split_and_process_bio from dm_request, we set | |
1579 | * DMF_QUEUE_IO_TO_THREAD. | |
1580 | * | |
1581 | * To quiesce the thread (dm_wq_work), we set DMF_BLOCK_IO_FOR_SUSPEND | |
1582 | * and call flush_workqueue(md->wq). flush_workqueue will wait until | |
1583 | * dm_wq_work exits and DMF_BLOCK_IO_FOR_SUSPEND will prevent any | |
1584 | * further calls to __split_and_process_bio from dm_wq_work. | |
1da177e4 | 1585 | */ |
2ca3310e | 1586 | down_write(&md->io_lock); |
1eb787ec AK |
1587 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1588 | set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); | |
2ca3310e | 1589 | up_write(&md->io_lock); |
1da177e4 | 1590 | |
3b00b203 MP |
1591 | flush_workqueue(md->wq); |
1592 | ||
1da177e4 | 1593 | /* |
3b00b203 MP |
1594 | * At this point no more requests are entering target request routines. |
1595 | * We call dm_wait_for_completion to wait for all existing requests | |
1596 | * to finish. | |
1da177e4 | 1597 | */ |
401600df | 1598 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 1599 | |
2ca3310e | 1600 | down_write(&md->io_lock); |
6d6f10df | 1601 | if (noflush) |
022c2611 | 1602 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 1603 | up_write(&md->io_lock); |
2e93ccc1 | 1604 | |
1da177e4 | 1605 | /* were we interrupted ? */ |
46125c1c | 1606 | if (r < 0) { |
9a1fb464 | 1607 | dm_queue_flush(md); |
73d410c0 | 1608 | |
2ca3310e | 1609 | unlock_fs(md); |
2e93ccc1 | 1610 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1611 | } |
1da177e4 | 1612 | |
3b00b203 MP |
1613 | /* |
1614 | * If dm_wait_for_completion returned 0, the device is completely | |
1615 | * quiescent now. There is no request-processing activity. All new | |
1616 | * requests are being added to md->deferred list. | |
1617 | */ | |
1618 | ||
cf222b37 | 1619 | dm_table_postsuspend_targets(map); |
1da177e4 | 1620 | |
2ca3310e | 1621 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1622 | |
2ca3310e | 1623 | out: |
e39e2e95 AK |
1624 | if (r && md->suspended_bdev) { |
1625 | bdput(md->suspended_bdev); | |
1626 | md->suspended_bdev = NULL; | |
1627 | } | |
1628 | ||
2ca3310e | 1629 | dm_table_put(map); |
d287483d AK |
1630 | |
1631 | out_unlock: | |
e61290a4 | 1632 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1633 | return r; |
1da177e4 LT |
1634 | } |
1635 | ||
1636 | int dm_resume(struct mapped_device *md) | |
1637 | { | |
cf222b37 | 1638 | int r = -EINVAL; |
cf222b37 | 1639 | struct dm_table *map = NULL; |
1da177e4 | 1640 | |
e61290a4 | 1641 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1642 | if (!dm_suspended(md)) |
cf222b37 | 1643 | goto out; |
cf222b37 AK |
1644 | |
1645 | map = dm_get_table(md); | |
2ca3310e | 1646 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1647 | goto out; |
1da177e4 | 1648 | |
8757b776 MB |
1649 | r = dm_table_resume_targets(map); |
1650 | if (r) | |
1651 | goto out; | |
2ca3310e | 1652 | |
9a1fb464 | 1653 | dm_queue_flush(md); |
2ca3310e AK |
1654 | |
1655 | unlock_fs(md); | |
1656 | ||
bfa152fa JN |
1657 | if (md->suspended_bdev) { |
1658 | bdput(md->suspended_bdev); | |
1659 | md->suspended_bdev = NULL; | |
1660 | } | |
e39e2e95 | 1661 | |
2ca3310e AK |
1662 | clear_bit(DMF_SUSPENDED, &md->flags); |
1663 | ||
1da177e4 | 1664 | dm_table_unplug_all(map); |
1da177e4 | 1665 | |
69267a30 | 1666 | dm_kobject_uevent(md); |
8560ed6f | 1667 | |
cf222b37 | 1668 | r = 0; |
2ca3310e | 1669 | |
cf222b37 AK |
1670 | out: |
1671 | dm_table_put(map); | |
e61290a4 | 1672 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1673 | |
cf222b37 | 1674 | return r; |
1da177e4 LT |
1675 | } |
1676 | ||
1677 | /*----------------------------------------------------------------- | |
1678 | * Event notification. | |
1679 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1680 | void dm_kobject_uevent(struct mapped_device *md) |
1681 | { | |
ed9e1982 | 1682 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1683 | } |
1684 | ||
7a8c3d3b MA |
1685 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1686 | { | |
1687 | return atomic_add_return(1, &md->uevent_seq); | |
1688 | } | |
1689 | ||
1da177e4 LT |
1690 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1691 | { | |
1692 | return atomic_read(&md->event_nr); | |
1693 | } | |
1694 | ||
1695 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1696 | { | |
1697 | return wait_event_interruptible(md->eventq, | |
1698 | (event_nr != atomic_read(&md->event_nr))); | |
1699 | } | |
1700 | ||
7a8c3d3b MA |
1701 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1702 | { | |
1703 | unsigned long flags; | |
1704 | ||
1705 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1706 | list_add(elist, &md->uevent_list); | |
1707 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1708 | } | |
1709 | ||
1da177e4 LT |
1710 | /* |
1711 | * The gendisk is only valid as long as you have a reference | |
1712 | * count on 'md'. | |
1713 | */ | |
1714 | struct gendisk *dm_disk(struct mapped_device *md) | |
1715 | { | |
1716 | return md->disk; | |
1717 | } | |
1718 | ||
784aae73 MB |
1719 | struct kobject *dm_kobject(struct mapped_device *md) |
1720 | { | |
1721 | return &md->kobj; | |
1722 | } | |
1723 | ||
1724 | /* | |
1725 | * struct mapped_device should not be exported outside of dm.c | |
1726 | * so use this check to verify that kobj is part of md structure | |
1727 | */ | |
1728 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1729 | { | |
1730 | struct mapped_device *md; | |
1731 | ||
1732 | md = container_of(kobj, struct mapped_device, kobj); | |
1733 | if (&md->kobj != kobj) | |
1734 | return NULL; | |
1735 | ||
1736 | dm_get(md); | |
1737 | return md; | |
1738 | } | |
1739 | ||
1da177e4 LT |
1740 | int dm_suspended(struct mapped_device *md) |
1741 | { | |
1742 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1743 | } | |
1744 | ||
2e93ccc1 KU |
1745 | int dm_noflush_suspending(struct dm_target *ti) |
1746 | { | |
1747 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1748 | int r = __noflush_suspending(md); | |
1749 | ||
1750 | dm_put(md); | |
1751 | ||
1752 | return r; | |
1753 | } | |
1754 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1755 | ||
1da177e4 LT |
1756 | static struct block_device_operations dm_blk_dops = { |
1757 | .open = dm_blk_open, | |
1758 | .release = dm_blk_close, | |
aa129a22 | 1759 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1760 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1761 | .owner = THIS_MODULE |
1762 | }; | |
1763 | ||
1764 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1765 | ||
1766 | /* | |
1767 | * module hooks | |
1768 | */ | |
1769 | module_init(dm_init); | |
1770 | module_exit(dm_exit); | |
1771 | ||
1772 | module_param(major, uint, 0); | |
1773 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1774 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1775 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1776 | MODULE_LICENSE("GPL"); |