]>
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 | |
af7e466a MP |
127 | /* |
128 | * An error from the barrier request currently being processed. | |
129 | */ | |
130 | int barrier_error; | |
131 | ||
304f3f6a MB |
132 | /* |
133 | * Processing queue (flush/barriers) | |
134 | */ | |
135 | struct workqueue_struct *wq; | |
136 | ||
1da177e4 LT |
137 | /* |
138 | * The current mapping. | |
139 | */ | |
140 | struct dm_table *map; | |
141 | ||
142 | /* | |
143 | * io objects are allocated from here. | |
144 | */ | |
145 | mempool_t *io_pool; | |
146 | mempool_t *tio_pool; | |
147 | ||
9faf400f SB |
148 | struct bio_set *bs; |
149 | ||
1da177e4 LT |
150 | /* |
151 | * Event handling. | |
152 | */ | |
153 | atomic_t event_nr; | |
154 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
155 | atomic_t uevent_seq; |
156 | struct list_head uevent_list; | |
157 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
158 | |
159 | /* | |
160 | * freeze/thaw support require holding onto a super block | |
161 | */ | |
162 | struct super_block *frozen_sb; | |
e39e2e95 | 163 | struct block_device *suspended_bdev; |
3ac51e74 DW |
164 | |
165 | /* forced geometry settings */ | |
166 | struct hd_geometry geometry; | |
784aae73 MB |
167 | |
168 | /* sysfs handle */ | |
169 | struct kobject kobj; | |
1da177e4 LT |
170 | }; |
171 | ||
172 | #define MIN_IOS 256 | |
e18b890b CL |
173 | static struct kmem_cache *_io_cache; |
174 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
175 | static struct kmem_cache *_rq_tio_cache; |
176 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 177 | |
1da177e4 LT |
178 | static int __init local_init(void) |
179 | { | |
51157b4a | 180 | int r = -ENOMEM; |
1da177e4 | 181 | |
1da177e4 | 182 | /* allocate a slab for the dm_ios */ |
028867ac | 183 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 184 | if (!_io_cache) |
51157b4a | 185 | return r; |
1da177e4 LT |
186 | |
187 | /* allocate a slab for the target ios */ | |
028867ac | 188 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
189 | if (!_tio_cache) |
190 | goto out_free_io_cache; | |
1da177e4 | 191 | |
8fbf26ad KU |
192 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
193 | if (!_rq_tio_cache) | |
194 | goto out_free_tio_cache; | |
195 | ||
196 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
197 | if (!_rq_bio_info_cache) | |
198 | goto out_free_rq_tio_cache; | |
199 | ||
51e5b2bd | 200 | r = dm_uevent_init(); |
51157b4a | 201 | if (r) |
8fbf26ad | 202 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 203 | |
1da177e4 LT |
204 | _major = major; |
205 | r = register_blkdev(_major, _name); | |
51157b4a KU |
206 | if (r < 0) |
207 | goto out_uevent_exit; | |
1da177e4 LT |
208 | |
209 | if (!_major) | |
210 | _major = r; | |
211 | ||
212 | return 0; | |
51157b4a KU |
213 | |
214 | out_uevent_exit: | |
215 | dm_uevent_exit(); | |
8fbf26ad KU |
216 | out_free_rq_bio_info_cache: |
217 | kmem_cache_destroy(_rq_bio_info_cache); | |
218 | out_free_rq_tio_cache: | |
219 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
220 | out_free_tio_cache: |
221 | kmem_cache_destroy(_tio_cache); | |
222 | out_free_io_cache: | |
223 | kmem_cache_destroy(_io_cache); | |
224 | ||
225 | return r; | |
1da177e4 LT |
226 | } |
227 | ||
228 | static void local_exit(void) | |
229 | { | |
8fbf26ad KU |
230 | kmem_cache_destroy(_rq_bio_info_cache); |
231 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
232 | kmem_cache_destroy(_tio_cache); |
233 | kmem_cache_destroy(_io_cache); | |
00d59405 | 234 | unregister_blkdev(_major, _name); |
51e5b2bd | 235 | dm_uevent_exit(); |
1da177e4 LT |
236 | |
237 | _major = 0; | |
238 | ||
239 | DMINFO("cleaned up"); | |
240 | } | |
241 | ||
b9249e55 | 242 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
243 | local_init, |
244 | dm_target_init, | |
245 | dm_linear_init, | |
246 | dm_stripe_init, | |
945fa4d2 | 247 | dm_kcopyd_init, |
1da177e4 LT |
248 | dm_interface_init, |
249 | }; | |
250 | ||
b9249e55 | 251 | static void (*_exits[])(void) = { |
1da177e4 LT |
252 | local_exit, |
253 | dm_target_exit, | |
254 | dm_linear_exit, | |
255 | dm_stripe_exit, | |
945fa4d2 | 256 | dm_kcopyd_exit, |
1da177e4 LT |
257 | dm_interface_exit, |
258 | }; | |
259 | ||
260 | static int __init dm_init(void) | |
261 | { | |
262 | const int count = ARRAY_SIZE(_inits); | |
263 | ||
264 | int r, i; | |
265 | ||
266 | for (i = 0; i < count; i++) { | |
267 | r = _inits[i](); | |
268 | if (r) | |
269 | goto bad; | |
270 | } | |
271 | ||
272 | return 0; | |
273 | ||
274 | bad: | |
275 | while (i--) | |
276 | _exits[i](); | |
277 | ||
278 | return r; | |
279 | } | |
280 | ||
281 | static void __exit dm_exit(void) | |
282 | { | |
283 | int i = ARRAY_SIZE(_exits); | |
284 | ||
285 | while (i--) | |
286 | _exits[i](); | |
287 | } | |
288 | ||
289 | /* | |
290 | * Block device functions | |
291 | */ | |
fe5f9f2c | 292 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
293 | { |
294 | struct mapped_device *md; | |
295 | ||
fba9f90e JM |
296 | spin_lock(&_minor_lock); |
297 | ||
fe5f9f2c | 298 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
299 | if (!md) |
300 | goto out; | |
301 | ||
5c6bd75d AK |
302 | if (test_bit(DMF_FREEING, &md->flags) || |
303 | test_bit(DMF_DELETING, &md->flags)) { | |
fba9f90e JM |
304 | md = NULL; |
305 | goto out; | |
306 | } | |
307 | ||
1da177e4 | 308 | dm_get(md); |
5c6bd75d | 309 | atomic_inc(&md->open_count); |
fba9f90e JM |
310 | |
311 | out: | |
312 | spin_unlock(&_minor_lock); | |
313 | ||
314 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
315 | } |
316 | ||
fe5f9f2c | 317 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 318 | { |
fe5f9f2c | 319 | struct mapped_device *md = disk->private_data; |
5c6bd75d | 320 | atomic_dec(&md->open_count); |
1da177e4 LT |
321 | dm_put(md); |
322 | return 0; | |
323 | } | |
324 | ||
5c6bd75d AK |
325 | int dm_open_count(struct mapped_device *md) |
326 | { | |
327 | return atomic_read(&md->open_count); | |
328 | } | |
329 | ||
330 | /* | |
331 | * Guarantees nothing is using the device before it's deleted. | |
332 | */ | |
333 | int dm_lock_for_deletion(struct mapped_device *md) | |
334 | { | |
335 | int r = 0; | |
336 | ||
337 | spin_lock(&_minor_lock); | |
338 | ||
339 | if (dm_open_count(md)) | |
340 | r = -EBUSY; | |
341 | else | |
342 | set_bit(DMF_DELETING, &md->flags); | |
343 | ||
344 | spin_unlock(&_minor_lock); | |
345 | ||
346 | return r; | |
347 | } | |
348 | ||
3ac51e74 DW |
349 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
350 | { | |
351 | struct mapped_device *md = bdev->bd_disk->private_data; | |
352 | ||
353 | return dm_get_geometry(md, geo); | |
354 | } | |
355 | ||
fe5f9f2c | 356 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
357 | unsigned int cmd, unsigned long arg) |
358 | { | |
fe5f9f2c AV |
359 | struct mapped_device *md = bdev->bd_disk->private_data; |
360 | struct dm_table *map = dm_get_table(md); | |
aa129a22 MB |
361 | struct dm_target *tgt; |
362 | int r = -ENOTTY; | |
363 | ||
aa129a22 MB |
364 | if (!map || !dm_table_get_size(map)) |
365 | goto out; | |
366 | ||
367 | /* We only support devices that have a single target */ | |
368 | if (dm_table_get_num_targets(map) != 1) | |
369 | goto out; | |
370 | ||
371 | tgt = dm_table_get_target(map, 0); | |
372 | ||
373 | if (dm_suspended(md)) { | |
374 | r = -EAGAIN; | |
375 | goto out; | |
376 | } | |
377 | ||
378 | if (tgt->type->ioctl) | |
647b3d00 | 379 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
380 | |
381 | out: | |
382 | dm_table_put(map); | |
383 | ||
aa129a22 MB |
384 | return r; |
385 | } | |
386 | ||
028867ac | 387 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
388 | { |
389 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
390 | } | |
391 | ||
028867ac | 392 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
393 | { |
394 | mempool_free(io, md->io_pool); | |
395 | } | |
396 | ||
028867ac | 397 | static struct dm_target_io *alloc_tio(struct mapped_device *md) |
1da177e4 LT |
398 | { |
399 | return mempool_alloc(md->tio_pool, GFP_NOIO); | |
400 | } | |
401 | ||
028867ac | 402 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
403 | { |
404 | mempool_free(tio, md->tio_pool); | |
405 | } | |
406 | ||
3eaf840e JNN |
407 | static void start_io_acct(struct dm_io *io) |
408 | { | |
409 | struct mapped_device *md = io->md; | |
c9959059 | 410 | int cpu; |
3eaf840e JNN |
411 | |
412 | io->start_time = jiffies; | |
413 | ||
074a7aca TH |
414 | cpu = part_stat_lock(); |
415 | part_round_stats(cpu, &dm_disk(md)->part0); | |
416 | part_stat_unlock(); | |
417 | dm_disk(md)->part0.in_flight = atomic_inc_return(&md->pending); | |
3eaf840e JNN |
418 | } |
419 | ||
d221d2e7 | 420 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
421 | { |
422 | struct mapped_device *md = io->md; | |
423 | struct bio *bio = io->bio; | |
424 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 425 | int pending, cpu; |
3eaf840e JNN |
426 | int rw = bio_data_dir(bio); |
427 | ||
074a7aca TH |
428 | cpu = part_stat_lock(); |
429 | part_round_stats(cpu, &dm_disk(md)->part0); | |
430 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
431 | part_stat_unlock(); | |
3eaf840e | 432 | |
af7e466a MP |
433 | /* |
434 | * After this is decremented the bio must not be touched if it is | |
435 | * a barrier. | |
436 | */ | |
074a7aca TH |
437 | dm_disk(md)->part0.in_flight = pending = |
438 | atomic_dec_return(&md->pending); | |
3eaf840e | 439 | |
d221d2e7 MP |
440 | /* nudge anyone waiting on suspend queue */ |
441 | if (!pending) | |
442 | wake_up(&md->wait); | |
3eaf840e JNN |
443 | } |
444 | ||
1da177e4 LT |
445 | /* |
446 | * Add the bio to the list of deferred io. | |
447 | */ | |
92c63902 | 448 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 449 | { |
2ca3310e | 450 | down_write(&md->io_lock); |
1da177e4 | 451 | |
022c2611 | 452 | spin_lock_irq(&md->deferred_lock); |
1da177e4 | 453 | bio_list_add(&md->deferred, bio); |
022c2611 | 454 | spin_unlock_irq(&md->deferred_lock); |
1da177e4 | 455 | |
92c63902 MP |
456 | if (!test_and_set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) |
457 | queue_work(md->wq, &md->work); | |
458 | ||
2ca3310e | 459 | up_write(&md->io_lock); |
1da177e4 LT |
460 | } |
461 | ||
462 | /* | |
463 | * Everyone (including functions in this file), should use this | |
464 | * function to access the md->map field, and make sure they call | |
465 | * dm_table_put() when finished. | |
466 | */ | |
467 | struct dm_table *dm_get_table(struct mapped_device *md) | |
468 | { | |
469 | struct dm_table *t; | |
470 | ||
471 | read_lock(&md->map_lock); | |
472 | t = md->map; | |
473 | if (t) | |
474 | dm_table_get(t); | |
475 | read_unlock(&md->map_lock); | |
476 | ||
477 | return t; | |
478 | } | |
479 | ||
3ac51e74 DW |
480 | /* |
481 | * Get the geometry associated with a dm device | |
482 | */ | |
483 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
484 | { | |
485 | *geo = md->geometry; | |
486 | ||
487 | return 0; | |
488 | } | |
489 | ||
490 | /* | |
491 | * Set the geometry of a device. | |
492 | */ | |
493 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
494 | { | |
495 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
496 | ||
497 | if (geo->start > sz) { | |
498 | DMWARN("Start sector is beyond the geometry limits."); | |
499 | return -EINVAL; | |
500 | } | |
501 | ||
502 | md->geometry = *geo; | |
503 | ||
504 | return 0; | |
505 | } | |
506 | ||
1da177e4 LT |
507 | /*----------------------------------------------------------------- |
508 | * CRUD START: | |
509 | * A more elegant soln is in the works that uses the queue | |
510 | * merge fn, unfortunately there are a couple of changes to | |
511 | * the block layer that I want to make for this. So in the | |
512 | * interests of getting something for people to use I give | |
513 | * you this clearly demarcated crap. | |
514 | *---------------------------------------------------------------*/ | |
515 | ||
2e93ccc1 KU |
516 | static int __noflush_suspending(struct mapped_device *md) |
517 | { | |
518 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
519 | } | |
520 | ||
1da177e4 LT |
521 | /* |
522 | * Decrements the number of outstanding ios that a bio has been | |
523 | * cloned into, completing the original io if necc. | |
524 | */ | |
858119e1 | 525 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 526 | { |
2e93ccc1 | 527 | unsigned long flags; |
b35f8caa MB |
528 | int io_error; |
529 | struct bio *bio; | |
530 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
531 | |
532 | /* Push-back supersedes any I/O errors */ | |
b35f8caa | 533 | if (error && !(io->error > 0 && __noflush_suspending(md))) |
1da177e4 LT |
534 | io->error = error; |
535 | ||
536 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
537 | if (io->error == DM_ENDIO_REQUEUE) { |
538 | /* | |
539 | * Target requested pushing back the I/O. | |
2e93ccc1 | 540 | */ |
022c2611 | 541 | spin_lock_irqsave(&md->deferred_lock, flags); |
b35f8caa | 542 | if (__noflush_suspending(md)) |
af7e466a | 543 | bio_list_add_head(&md->deferred, io->bio); |
2e93ccc1 KU |
544 | else |
545 | /* noflush suspend was interrupted. */ | |
546 | io->error = -EIO; | |
022c2611 | 547 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
548 | } |
549 | ||
b35f8caa MB |
550 | io_error = io->error; |
551 | bio = io->bio; | |
2e93ccc1 | 552 | |
af7e466a MP |
553 | if (bio_barrier(bio)) { |
554 | /* | |
555 | * There can be just one barrier request so we use | |
556 | * a per-device variable for error reporting. | |
557 | * Note that you can't touch the bio after end_io_acct | |
558 | */ | |
559 | md->barrier_error = io_error; | |
560 | end_io_acct(io); | |
561 | } else { | |
562 | end_io_acct(io); | |
b35f8caa | 563 | |
af7e466a MP |
564 | if (io_error != DM_ENDIO_REQUEUE) { |
565 | trace_block_bio_complete(md->queue, bio); | |
2056a782 | 566 | |
af7e466a MP |
567 | bio_endio(bio, io_error); |
568 | } | |
b35f8caa | 569 | } |
af7e466a MP |
570 | |
571 | free_io(md, io); | |
1da177e4 LT |
572 | } |
573 | } | |
574 | ||
6712ecf8 | 575 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
576 | { |
577 | int r = 0; | |
028867ac | 578 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 579 | struct dm_io *io = tio->io; |
9faf400f | 580 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
581 | dm_endio_fn endio = tio->ti->type->end_io; |
582 | ||
1da177e4 LT |
583 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
584 | error = -EIO; | |
585 | ||
586 | if (endio) { | |
587 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
588 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
589 | /* | |
590 | * error and requeue request are handled | |
591 | * in dec_pending(). | |
592 | */ | |
1da177e4 | 593 | error = r; |
45cbcd79 KU |
594 | else if (r == DM_ENDIO_INCOMPLETE) |
595 | /* The target will handle the io */ | |
6712ecf8 | 596 | return; |
45cbcd79 KU |
597 | else if (r) { |
598 | DMWARN("unimplemented target endio return value: %d", r); | |
599 | BUG(); | |
600 | } | |
1da177e4 LT |
601 | } |
602 | ||
9faf400f SB |
603 | /* |
604 | * Store md for cleanup instead of tio which is about to get freed. | |
605 | */ | |
606 | bio->bi_private = md->bs; | |
607 | ||
9faf400f | 608 | free_tio(md, tio); |
b35f8caa MB |
609 | bio_put(bio); |
610 | dec_pending(io, error); | |
1da177e4 LT |
611 | } |
612 | ||
613 | static sector_t max_io_len(struct mapped_device *md, | |
614 | sector_t sector, struct dm_target *ti) | |
615 | { | |
616 | sector_t offset = sector - ti->begin; | |
617 | sector_t len = ti->len - offset; | |
618 | ||
619 | /* | |
620 | * Does the target need to split even further ? | |
621 | */ | |
622 | if (ti->split_io) { | |
623 | sector_t boundary; | |
624 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) | |
625 | - offset; | |
626 | if (len > boundary) | |
627 | len = boundary; | |
628 | } | |
629 | ||
630 | return len; | |
631 | } | |
632 | ||
633 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 634 | struct dm_target_io *tio) |
1da177e4 LT |
635 | { |
636 | int r; | |
2056a782 | 637 | sector_t sector; |
9faf400f | 638 | struct mapped_device *md; |
1da177e4 LT |
639 | |
640 | /* | |
641 | * Sanity checks. | |
642 | */ | |
643 | BUG_ON(!clone->bi_size); | |
644 | ||
645 | clone->bi_end_io = clone_endio; | |
646 | clone->bi_private = tio; | |
647 | ||
648 | /* | |
649 | * Map the clone. If r == 0 we don't need to do | |
650 | * anything, the target has assumed ownership of | |
651 | * this io. | |
652 | */ | |
653 | atomic_inc(&tio->io->io_count); | |
2056a782 | 654 | sector = clone->bi_sector; |
1da177e4 | 655 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 656 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 657 | /* the bio has been remapped so dispatch it */ |
2056a782 | 658 | |
5f3ea37c | 659 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
c7149d6b AB |
660 | tio->io->bio->bi_bdev->bd_dev, |
661 | clone->bi_sector, sector); | |
2056a782 | 662 | |
1da177e4 | 663 | generic_make_request(clone); |
2e93ccc1 KU |
664 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
665 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
666 | md = tio->io->md; |
667 | dec_pending(tio->io, r); | |
668 | /* | |
669 | * Store bio_set for cleanup. | |
670 | */ | |
671 | clone->bi_private = md->bs; | |
1da177e4 | 672 | bio_put(clone); |
9faf400f | 673 | free_tio(md, tio); |
45cbcd79 KU |
674 | } else if (r) { |
675 | DMWARN("unimplemented target map return value: %d", r); | |
676 | BUG(); | |
1da177e4 LT |
677 | } |
678 | } | |
679 | ||
680 | struct clone_info { | |
681 | struct mapped_device *md; | |
682 | struct dm_table *map; | |
683 | struct bio *bio; | |
684 | struct dm_io *io; | |
685 | sector_t sector; | |
686 | sector_t sector_count; | |
687 | unsigned short idx; | |
688 | }; | |
689 | ||
3676347a PO |
690 | static void dm_bio_destructor(struct bio *bio) |
691 | { | |
9faf400f SB |
692 | struct bio_set *bs = bio->bi_private; |
693 | ||
694 | bio_free(bio, bs); | |
3676347a PO |
695 | } |
696 | ||
1da177e4 LT |
697 | /* |
698 | * Creates a little bio that is just does part of a bvec. | |
699 | */ | |
700 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
701 | unsigned short idx, unsigned int offset, | |
9faf400f | 702 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
703 | { |
704 | struct bio *clone; | |
705 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
706 | ||
9faf400f | 707 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 708 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
709 | *clone->bi_io_vec = *bv; |
710 | ||
711 | clone->bi_sector = sector; | |
712 | clone->bi_bdev = bio->bi_bdev; | |
af7e466a | 713 | clone->bi_rw = bio->bi_rw & ~(1 << BIO_RW_BARRIER); |
1da177e4 LT |
714 | clone->bi_vcnt = 1; |
715 | clone->bi_size = to_bytes(len); | |
716 | clone->bi_io_vec->bv_offset = offset; | |
717 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 718 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 719 | |
9c47008d MP |
720 | if (bio_integrity(bio)) { |
721 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
722 | bio_integrity_trim(clone, | |
723 | bio_sector_offset(bio, idx, offset), len); | |
724 | } | |
725 | ||
1da177e4 LT |
726 | return clone; |
727 | } | |
728 | ||
729 | /* | |
730 | * Creates a bio that consists of range of complete bvecs. | |
731 | */ | |
732 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
733 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 734 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
735 | { |
736 | struct bio *clone; | |
737 | ||
9faf400f SB |
738 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
739 | __bio_clone(clone, bio); | |
af7e466a | 740 | clone->bi_rw &= ~(1 << BIO_RW_BARRIER); |
9faf400f | 741 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
742 | clone->bi_sector = sector; |
743 | clone->bi_idx = idx; | |
744 | clone->bi_vcnt = idx + bv_count; | |
745 | clone->bi_size = to_bytes(len); | |
746 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
747 | ||
9c47008d MP |
748 | if (bio_integrity(bio)) { |
749 | bio_integrity_clone(clone, bio, GFP_NOIO); | |
750 | ||
751 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
752 | bio_integrity_trim(clone, | |
753 | bio_sector_offset(bio, idx, 0), len); | |
754 | } | |
755 | ||
1da177e4 LT |
756 | return clone; |
757 | } | |
758 | ||
512875bd | 759 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
760 | { |
761 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
762 | struct dm_target *ti; |
763 | sector_t len = 0, max; | |
028867ac | 764 | struct dm_target_io *tio; |
1da177e4 | 765 | |
512875bd JN |
766 | ti = dm_table_find_target(ci->map, ci->sector); |
767 | if (!dm_target_is_valid(ti)) | |
768 | return -EIO; | |
769 | ||
770 | max = max_io_len(ci->md, ci->sector, ti); | |
771 | ||
1da177e4 LT |
772 | /* |
773 | * Allocate a target io object. | |
774 | */ | |
775 | tio = alloc_tio(ci->md); | |
776 | tio->io = ci->io; | |
777 | tio->ti = ti; | |
778 | memset(&tio->info, 0, sizeof(tio->info)); | |
779 | ||
780 | if (ci->sector_count <= max) { | |
781 | /* | |
782 | * Optimise for the simple case where we can do all of | |
783 | * the remaining io with a single clone. | |
784 | */ | |
785 | clone = clone_bio(bio, ci->sector, ci->idx, | |
9faf400f SB |
786 | bio->bi_vcnt - ci->idx, ci->sector_count, |
787 | ci->md->bs); | |
1da177e4 LT |
788 | __map_bio(ti, clone, tio); |
789 | ci->sector_count = 0; | |
790 | ||
791 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
792 | /* | |
793 | * There are some bvecs that don't span targets. | |
794 | * Do as many of these as possible. | |
795 | */ | |
796 | int i; | |
797 | sector_t remaining = max; | |
798 | sector_t bv_len; | |
799 | ||
800 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
801 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
802 | ||
803 | if (bv_len > remaining) | |
804 | break; | |
805 | ||
806 | remaining -= bv_len; | |
807 | len += bv_len; | |
808 | } | |
809 | ||
9faf400f SB |
810 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
811 | ci->md->bs); | |
1da177e4 LT |
812 | __map_bio(ti, clone, tio); |
813 | ||
814 | ci->sector += len; | |
815 | ci->sector_count -= len; | |
816 | ci->idx = i; | |
817 | ||
818 | } else { | |
819 | /* | |
d2044a94 | 820 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
821 | */ |
822 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
823 | sector_t remaining = to_sector(bv->bv_len); |
824 | unsigned int offset = 0; | |
1da177e4 | 825 | |
d2044a94 AK |
826 | do { |
827 | if (offset) { | |
828 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
829 | if (!dm_target_is_valid(ti)) |
830 | return -EIO; | |
831 | ||
d2044a94 | 832 | max = max_io_len(ci->md, ci->sector, ti); |
1da177e4 | 833 | |
d2044a94 AK |
834 | tio = alloc_tio(ci->md); |
835 | tio->io = ci->io; | |
836 | tio->ti = ti; | |
837 | memset(&tio->info, 0, sizeof(tio->info)); | |
838 | } | |
839 | ||
840 | len = min(remaining, max); | |
841 | ||
842 | clone = split_bvec(bio, ci->sector, ci->idx, | |
9faf400f SB |
843 | bv->bv_offset + offset, len, |
844 | ci->md->bs); | |
d2044a94 AK |
845 | |
846 | __map_bio(ti, clone, tio); | |
847 | ||
848 | ci->sector += len; | |
849 | ci->sector_count -= len; | |
850 | offset += to_bytes(len); | |
851 | } while (remaining -= len); | |
1da177e4 | 852 | |
1da177e4 LT |
853 | ci->idx++; |
854 | } | |
512875bd JN |
855 | |
856 | return 0; | |
1da177e4 LT |
857 | } |
858 | ||
859 | /* | |
8a53c28d | 860 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 861 | */ |
f0b9a450 | 862 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
863 | { |
864 | struct clone_info ci; | |
512875bd | 865 | int error = 0; |
1da177e4 LT |
866 | |
867 | ci.map = dm_get_table(md); | |
f0b9a450 | 868 | if (unlikely(!ci.map)) { |
af7e466a MP |
869 | if (!bio_barrier(bio)) |
870 | bio_io_error(bio); | |
871 | else | |
872 | md->barrier_error = -EIO; | |
f0b9a450 MP |
873 | return; |
874 | } | |
692d0eb9 | 875 | |
1da177e4 LT |
876 | ci.md = md; |
877 | ci.bio = bio; | |
878 | ci.io = alloc_io(md); | |
879 | ci.io->error = 0; | |
880 | atomic_set(&ci.io->io_count, 1); | |
881 | ci.io->bio = bio; | |
882 | ci.io->md = md; | |
883 | ci.sector = bio->bi_sector; | |
884 | ci.sector_count = bio_sectors(bio); | |
885 | ci.idx = bio->bi_idx; | |
886 | ||
3eaf840e | 887 | start_io_acct(ci.io); |
512875bd JN |
888 | while (ci.sector_count && !error) |
889 | error = __clone_and_map(&ci); | |
1da177e4 LT |
890 | |
891 | /* drop the extra reference count */ | |
512875bd | 892 | dec_pending(ci.io, error); |
1da177e4 LT |
893 | dm_table_put(ci.map); |
894 | } | |
895 | /*----------------------------------------------------------------- | |
896 | * CRUD END | |
897 | *---------------------------------------------------------------*/ | |
898 | ||
f6fccb12 MB |
899 | static int dm_merge_bvec(struct request_queue *q, |
900 | struct bvec_merge_data *bvm, | |
901 | struct bio_vec *biovec) | |
902 | { | |
903 | struct mapped_device *md = q->queuedata; | |
904 | struct dm_table *map = dm_get_table(md); | |
905 | struct dm_target *ti; | |
906 | sector_t max_sectors; | |
5037108a | 907 | int max_size = 0; |
f6fccb12 MB |
908 | |
909 | if (unlikely(!map)) | |
5037108a | 910 | goto out; |
f6fccb12 MB |
911 | |
912 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
913 | if (!dm_target_is_valid(ti)) |
914 | goto out_table; | |
f6fccb12 MB |
915 | |
916 | /* | |
917 | * Find maximum amount of I/O that won't need splitting | |
918 | */ | |
919 | max_sectors = min(max_io_len(md, bvm->bi_sector, ti), | |
920 | (sector_t) BIO_MAX_SECTORS); | |
921 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
922 | if (max_size < 0) | |
923 | max_size = 0; | |
924 | ||
925 | /* | |
926 | * merge_bvec_fn() returns number of bytes | |
927 | * it can accept at this offset | |
928 | * max is precomputed maximal io size | |
929 | */ | |
930 | if (max_size && ti->type->merge) | |
931 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
932 | ||
b01cd5ac | 933 | out_table: |
5037108a MP |
934 | dm_table_put(map); |
935 | ||
936 | out: | |
f6fccb12 MB |
937 | /* |
938 | * Always allow an entire first page | |
939 | */ | |
940 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
941 | max_size = biovec->bv_len; | |
942 | ||
f6fccb12 MB |
943 | return max_size; |
944 | } | |
945 | ||
1da177e4 LT |
946 | /* |
947 | * The request function that just remaps the bio built up by | |
948 | * dm_merge_bvec. | |
949 | */ | |
165125e1 | 950 | static int dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 951 | { |
12f03a49 | 952 | int rw = bio_data_dir(bio); |
1da177e4 | 953 | struct mapped_device *md = q->queuedata; |
c9959059 | 954 | int cpu; |
1da177e4 | 955 | |
2ca3310e | 956 | down_read(&md->io_lock); |
1da177e4 | 957 | |
074a7aca TH |
958 | cpu = part_stat_lock(); |
959 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
960 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
961 | part_stat_unlock(); | |
12f03a49 | 962 | |
1da177e4 | 963 | /* |
1eb787ec AK |
964 | * If we're suspended or the thread is processing barriers |
965 | * we have to queue this io for later. | |
1da177e4 | 966 | */ |
af7e466a MP |
967 | if (unlikely(test_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) || |
968 | unlikely(bio_barrier(bio))) { | |
2ca3310e | 969 | up_read(&md->io_lock); |
1da177e4 | 970 | |
54d9a1b4 AK |
971 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) && |
972 | bio_rw(bio) == READA) { | |
973 | bio_io_error(bio); | |
974 | return 0; | |
975 | } | |
1da177e4 | 976 | |
92c63902 | 977 | queue_io(md, bio); |
1da177e4 | 978 | |
92c63902 | 979 | return 0; |
1da177e4 LT |
980 | } |
981 | ||
f0b9a450 | 982 | __split_and_process_bio(md, bio); |
2ca3310e | 983 | up_read(&md->io_lock); |
f0b9a450 | 984 | return 0; |
1da177e4 LT |
985 | } |
986 | ||
165125e1 | 987 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
988 | { |
989 | struct mapped_device *md = q->queuedata; | |
990 | struct dm_table *map = dm_get_table(md); | |
991 | ||
992 | if (map) { | |
993 | dm_table_unplug_all(map); | |
994 | dm_table_put(map); | |
995 | } | |
996 | } | |
997 | ||
998 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
999 | { | |
8a57dfc6 CS |
1000 | int r = bdi_bits; |
1001 | struct mapped_device *md = congested_data; | |
1002 | struct dm_table *map; | |
1da177e4 | 1003 | |
1eb787ec | 1004 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
8a57dfc6 CS |
1005 | map = dm_get_table(md); |
1006 | if (map) { | |
1007 | r = dm_table_any_congested(map, bdi_bits); | |
1008 | dm_table_put(map); | |
1009 | } | |
1010 | } | |
1011 | ||
1da177e4 LT |
1012 | return r; |
1013 | } | |
1014 | ||
1015 | /*----------------------------------------------------------------- | |
1016 | * An IDR is used to keep track of allocated minor numbers. | |
1017 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1018 | static DEFINE_IDR(_minor_idr); |
1019 | ||
2b06cfff | 1020 | static void free_minor(int minor) |
1da177e4 | 1021 | { |
f32c10b0 | 1022 | spin_lock(&_minor_lock); |
1da177e4 | 1023 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1024 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1025 | } |
1026 | ||
1027 | /* | |
1028 | * See if the device with a specific minor # is free. | |
1029 | */ | |
cf13ab8e | 1030 | static int specific_minor(int minor) |
1da177e4 LT |
1031 | { |
1032 | int r, m; | |
1033 | ||
1034 | if (minor >= (1 << MINORBITS)) | |
1035 | return -EINVAL; | |
1036 | ||
62f75c2f JM |
1037 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1038 | if (!r) | |
1039 | return -ENOMEM; | |
1040 | ||
f32c10b0 | 1041 | spin_lock(&_minor_lock); |
1da177e4 LT |
1042 | |
1043 | if (idr_find(&_minor_idr, minor)) { | |
1044 | r = -EBUSY; | |
1045 | goto out; | |
1046 | } | |
1047 | ||
ba61fdd1 | 1048 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1049 | if (r) |
1da177e4 | 1050 | goto out; |
1da177e4 LT |
1051 | |
1052 | if (m != minor) { | |
1053 | idr_remove(&_minor_idr, m); | |
1054 | r = -EBUSY; | |
1055 | goto out; | |
1056 | } | |
1057 | ||
1058 | out: | |
f32c10b0 | 1059 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1060 | return r; |
1061 | } | |
1062 | ||
cf13ab8e | 1063 | static int next_free_minor(int *minor) |
1da177e4 | 1064 | { |
2b06cfff | 1065 | int r, m; |
1da177e4 | 1066 | |
1da177e4 | 1067 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1068 | if (!r) |
1069 | return -ENOMEM; | |
1070 | ||
f32c10b0 | 1071 | spin_lock(&_minor_lock); |
1da177e4 | 1072 | |
ba61fdd1 | 1073 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1074 | if (r) |
1da177e4 | 1075 | goto out; |
1da177e4 LT |
1076 | |
1077 | if (m >= (1 << MINORBITS)) { | |
1078 | idr_remove(&_minor_idr, m); | |
1079 | r = -ENOSPC; | |
1080 | goto out; | |
1081 | } | |
1082 | ||
1083 | *minor = m; | |
1084 | ||
1085 | out: | |
f32c10b0 | 1086 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1087 | return r; |
1088 | } | |
1089 | ||
1090 | static struct block_device_operations dm_blk_dops; | |
1091 | ||
53d5914f MP |
1092 | static void dm_wq_work(struct work_struct *work); |
1093 | ||
1da177e4 LT |
1094 | /* |
1095 | * Allocate and initialise a blank device with a given minor. | |
1096 | */ | |
2b06cfff | 1097 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1098 | { |
1099 | int r; | |
cf13ab8e | 1100 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1101 | void *old_md; |
1da177e4 LT |
1102 | |
1103 | if (!md) { | |
1104 | DMWARN("unable to allocate device, out of memory."); | |
1105 | return NULL; | |
1106 | } | |
1107 | ||
10da4f79 | 1108 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1109 | goto bad_module_get; |
10da4f79 | 1110 | |
1da177e4 | 1111 | /* get a minor number for the dev */ |
2b06cfff | 1112 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1113 | r = next_free_minor(&minor); |
2b06cfff | 1114 | else |
cf13ab8e | 1115 | r = specific_minor(minor); |
1da177e4 | 1116 | if (r < 0) |
6ed7ade8 | 1117 | goto bad_minor; |
1da177e4 | 1118 | |
2ca3310e | 1119 | init_rwsem(&md->io_lock); |
e61290a4 | 1120 | mutex_init(&md->suspend_lock); |
022c2611 | 1121 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1122 | rwlock_init(&md->map_lock); |
1123 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1124 | atomic_set(&md->open_count, 0); |
1da177e4 | 1125 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1126 | atomic_set(&md->uevent_seq, 0); |
1127 | INIT_LIST_HEAD(&md->uevent_list); | |
1128 | spin_lock_init(&md->uevent_lock); | |
1da177e4 LT |
1129 | |
1130 | md->queue = blk_alloc_queue(GFP_KERNEL); | |
1131 | if (!md->queue) | |
6ed7ade8 | 1132 | goto bad_queue; |
1da177e4 LT |
1133 | |
1134 | md->queue->queuedata = md; | |
1135 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1136 | md->queue->backing_dev_info.congested_data = md; | |
1137 | blk_queue_make_request(md->queue, dm_request); | |
99360b4c | 1138 | blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN, NULL); |
daef265f | 1139 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); |
1da177e4 | 1140 | md->queue->unplug_fn = dm_unplug_all; |
f6fccb12 | 1141 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
1da177e4 | 1142 | |
93d2341c | 1143 | md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache); |
74859364 | 1144 | if (!md->io_pool) |
6ed7ade8 | 1145 | goto bad_io_pool; |
1da177e4 | 1146 | |
93d2341c | 1147 | md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache); |
1da177e4 | 1148 | if (!md->tio_pool) |
6ed7ade8 | 1149 | goto bad_tio_pool; |
1da177e4 | 1150 | |
bb799ca0 | 1151 | md->bs = bioset_create(16, 0); |
9faf400f SB |
1152 | if (!md->bs) |
1153 | goto bad_no_bioset; | |
1154 | ||
1da177e4 LT |
1155 | md->disk = alloc_disk(1); |
1156 | if (!md->disk) | |
6ed7ade8 | 1157 | goto bad_disk; |
1da177e4 | 1158 | |
f0b04115 JM |
1159 | atomic_set(&md->pending, 0); |
1160 | init_waitqueue_head(&md->wait); | |
53d5914f | 1161 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1162 | init_waitqueue_head(&md->eventq); |
1163 | ||
1da177e4 LT |
1164 | md->disk->major = _major; |
1165 | md->disk->first_minor = minor; | |
1166 | md->disk->fops = &dm_blk_dops; | |
1167 | md->disk->queue = md->queue; | |
1168 | md->disk->private_data = md; | |
1169 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1170 | add_disk(md->disk); | |
7e51f257 | 1171 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1172 | |
304f3f6a MB |
1173 | md->wq = create_singlethread_workqueue("kdmflush"); |
1174 | if (!md->wq) | |
1175 | goto bad_thread; | |
1176 | ||
ba61fdd1 | 1177 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1178 | spin_lock(&_minor_lock); |
ba61fdd1 | 1179 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1180 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1181 | |
1182 | BUG_ON(old_md != MINOR_ALLOCED); | |
1183 | ||
1da177e4 LT |
1184 | return md; |
1185 | ||
304f3f6a MB |
1186 | bad_thread: |
1187 | put_disk(md->disk); | |
6ed7ade8 | 1188 | bad_disk: |
9faf400f | 1189 | bioset_free(md->bs); |
6ed7ade8 | 1190 | bad_no_bioset: |
1da177e4 | 1191 | mempool_destroy(md->tio_pool); |
6ed7ade8 | 1192 | bad_tio_pool: |
1da177e4 | 1193 | mempool_destroy(md->io_pool); |
6ed7ade8 | 1194 | bad_io_pool: |
1312f40e | 1195 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1196 | bad_queue: |
1da177e4 | 1197 | free_minor(minor); |
6ed7ade8 | 1198 | bad_minor: |
10da4f79 | 1199 | module_put(THIS_MODULE); |
6ed7ade8 | 1200 | bad_module_get: |
1da177e4 LT |
1201 | kfree(md); |
1202 | return NULL; | |
1203 | } | |
1204 | ||
ae9da83f JN |
1205 | static void unlock_fs(struct mapped_device *md); |
1206 | ||
1da177e4 LT |
1207 | static void free_dev(struct mapped_device *md) |
1208 | { | |
f331c029 | 1209 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1210 | |
d9dde59b | 1211 | if (md->suspended_bdev) { |
ae9da83f | 1212 | unlock_fs(md); |
d9dde59b JN |
1213 | bdput(md->suspended_bdev); |
1214 | } | |
304f3f6a | 1215 | destroy_workqueue(md->wq); |
1da177e4 LT |
1216 | mempool_destroy(md->tio_pool); |
1217 | mempool_destroy(md->io_pool); | |
9faf400f | 1218 | bioset_free(md->bs); |
9c47008d | 1219 | blk_integrity_unregister(md->disk); |
1da177e4 | 1220 | del_gendisk(md->disk); |
63d94e48 | 1221 | free_minor(minor); |
fba9f90e JM |
1222 | |
1223 | spin_lock(&_minor_lock); | |
1224 | md->disk->private_data = NULL; | |
1225 | spin_unlock(&_minor_lock); | |
1226 | ||
1da177e4 | 1227 | put_disk(md->disk); |
1312f40e | 1228 | blk_cleanup_queue(md->queue); |
10da4f79 | 1229 | module_put(THIS_MODULE); |
1da177e4 LT |
1230 | kfree(md); |
1231 | } | |
1232 | ||
1233 | /* | |
1234 | * Bind a table to the device. | |
1235 | */ | |
1236 | static void event_callback(void *context) | |
1237 | { | |
7a8c3d3b MA |
1238 | unsigned long flags; |
1239 | LIST_HEAD(uevents); | |
1da177e4 LT |
1240 | struct mapped_device *md = (struct mapped_device *) context; |
1241 | ||
7a8c3d3b MA |
1242 | spin_lock_irqsave(&md->uevent_lock, flags); |
1243 | list_splice_init(&md->uevent_list, &uevents); | |
1244 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1245 | ||
ed9e1982 | 1246 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1247 | |
1da177e4 LT |
1248 | atomic_inc(&md->event_nr); |
1249 | wake_up(&md->eventq); | |
1250 | } | |
1251 | ||
4e90188b | 1252 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1253 | { |
4e90188b | 1254 | set_capacity(md->disk, size); |
1da177e4 | 1255 | |
1b1dcc1b | 1256 | mutex_lock(&md->suspended_bdev->bd_inode->i_mutex); |
e39e2e95 | 1257 | i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1b1dcc1b | 1258 | mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex); |
1da177e4 LT |
1259 | } |
1260 | ||
1261 | static int __bind(struct mapped_device *md, struct dm_table *t) | |
1262 | { | |
165125e1 | 1263 | struct request_queue *q = md->queue; |
1da177e4 LT |
1264 | sector_t size; |
1265 | ||
1266 | size = dm_table_get_size(t); | |
3ac51e74 DW |
1267 | |
1268 | /* | |
1269 | * Wipe any geometry if the size of the table changed. | |
1270 | */ | |
1271 | if (size != get_capacity(md->disk)) | |
1272 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
1273 | ||
bfa152fa JN |
1274 | if (md->suspended_bdev) |
1275 | __set_size(md, size); | |
d5816876 MP |
1276 | |
1277 | if (!size) { | |
1278 | dm_table_destroy(t); | |
1da177e4 | 1279 | return 0; |
d5816876 | 1280 | } |
1da177e4 | 1281 | |
2ca3310e AK |
1282 | dm_table_event_callback(t, event_callback, md); |
1283 | ||
1da177e4 LT |
1284 | write_lock(&md->map_lock); |
1285 | md->map = t; | |
2ca3310e | 1286 | dm_table_set_restrictions(t, q); |
1da177e4 LT |
1287 | write_unlock(&md->map_lock); |
1288 | ||
1da177e4 LT |
1289 | return 0; |
1290 | } | |
1291 | ||
1292 | static void __unbind(struct mapped_device *md) | |
1293 | { | |
1294 | struct dm_table *map = md->map; | |
1295 | ||
1296 | if (!map) | |
1297 | return; | |
1298 | ||
1299 | dm_table_event_callback(map, NULL, NULL); | |
1300 | write_lock(&md->map_lock); | |
1301 | md->map = NULL; | |
1302 | write_unlock(&md->map_lock); | |
d5816876 | 1303 | dm_table_destroy(map); |
1da177e4 LT |
1304 | } |
1305 | ||
1306 | /* | |
1307 | * Constructor for a new device. | |
1308 | */ | |
2b06cfff | 1309 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
1310 | { |
1311 | struct mapped_device *md; | |
1312 | ||
2b06cfff | 1313 | md = alloc_dev(minor); |
1da177e4 LT |
1314 | if (!md) |
1315 | return -ENXIO; | |
1316 | ||
784aae73 MB |
1317 | dm_sysfs_init(md); |
1318 | ||
1da177e4 LT |
1319 | *result = md; |
1320 | return 0; | |
1321 | } | |
1322 | ||
637842cf | 1323 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
1324 | { |
1325 | struct mapped_device *md; | |
1da177e4 LT |
1326 | unsigned minor = MINOR(dev); |
1327 | ||
1328 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
1329 | return NULL; | |
1330 | ||
f32c10b0 | 1331 | spin_lock(&_minor_lock); |
1da177e4 LT |
1332 | |
1333 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 1334 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 1335 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
17b2f66f | 1336 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 1337 | md = NULL; |
fba9f90e JM |
1338 | goto out; |
1339 | } | |
1da177e4 | 1340 | |
fba9f90e | 1341 | out: |
f32c10b0 | 1342 | spin_unlock(&_minor_lock); |
1da177e4 | 1343 | |
637842cf DT |
1344 | return md; |
1345 | } | |
1346 | ||
d229a958 DT |
1347 | struct mapped_device *dm_get_md(dev_t dev) |
1348 | { | |
1349 | struct mapped_device *md = dm_find_md(dev); | |
1350 | ||
1351 | if (md) | |
1352 | dm_get(md); | |
1353 | ||
1354 | return md; | |
1355 | } | |
1356 | ||
9ade92a9 | 1357 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 1358 | { |
9ade92a9 | 1359 | return md->interface_ptr; |
1da177e4 LT |
1360 | } |
1361 | ||
1362 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
1363 | { | |
1364 | md->interface_ptr = ptr; | |
1365 | } | |
1366 | ||
1367 | void dm_get(struct mapped_device *md) | |
1368 | { | |
1369 | atomic_inc(&md->holders); | |
1370 | } | |
1371 | ||
72d94861 AK |
1372 | const char *dm_device_name(struct mapped_device *md) |
1373 | { | |
1374 | return md->name; | |
1375 | } | |
1376 | EXPORT_SYMBOL_GPL(dm_device_name); | |
1377 | ||
1da177e4 LT |
1378 | void dm_put(struct mapped_device *md) |
1379 | { | |
1134e5ae | 1380 | struct dm_table *map; |
1da177e4 | 1381 | |
fba9f90e JM |
1382 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1383 | ||
f32c10b0 | 1384 | if (atomic_dec_and_lock(&md->holders, &_minor_lock)) { |
1134e5ae | 1385 | map = dm_get_table(md); |
f331c029 TH |
1386 | idr_replace(&_minor_idr, MINOR_ALLOCED, |
1387 | MINOR(disk_devt(dm_disk(md)))); | |
fba9f90e | 1388 | set_bit(DMF_FREEING, &md->flags); |
f32c10b0 | 1389 | spin_unlock(&_minor_lock); |
cf222b37 | 1390 | if (!dm_suspended(md)) { |
1da177e4 LT |
1391 | dm_table_presuspend_targets(map); |
1392 | dm_table_postsuspend_targets(map); | |
1393 | } | |
784aae73 | 1394 | dm_sysfs_exit(md); |
1134e5ae | 1395 | dm_table_put(map); |
a1b51e98 | 1396 | __unbind(md); |
1da177e4 LT |
1397 | free_dev(md); |
1398 | } | |
1da177e4 | 1399 | } |
79eb885c | 1400 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 1401 | |
401600df | 1402 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
1403 | { |
1404 | int r = 0; | |
b44ebeb0 MP |
1405 | DECLARE_WAITQUEUE(wait, current); |
1406 | ||
1407 | dm_unplug_all(md->queue); | |
1408 | ||
1409 | add_wait_queue(&md->wait, &wait); | |
46125c1c MB |
1410 | |
1411 | while (1) { | |
401600df | 1412 | set_current_state(interruptible); |
46125c1c MB |
1413 | |
1414 | smp_mb(); | |
1415 | if (!atomic_read(&md->pending)) | |
1416 | break; | |
1417 | ||
401600df MP |
1418 | if (interruptible == TASK_INTERRUPTIBLE && |
1419 | signal_pending(current)) { | |
46125c1c MB |
1420 | r = -EINTR; |
1421 | break; | |
1422 | } | |
1423 | ||
1424 | io_schedule(); | |
1425 | } | |
1426 | set_current_state(TASK_RUNNING); | |
1427 | ||
b44ebeb0 MP |
1428 | remove_wait_queue(&md->wait, &wait); |
1429 | ||
46125c1c MB |
1430 | return r; |
1431 | } | |
1432 | ||
af7e466a MP |
1433 | static int dm_flush(struct mapped_device *md) |
1434 | { | |
1435 | dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE); | |
1436 | return 0; | |
1437 | } | |
1438 | ||
1439 | static void process_barrier(struct mapped_device *md, struct bio *bio) | |
1440 | { | |
1441 | int error = dm_flush(md); | |
1442 | ||
1443 | if (unlikely(error)) { | |
1444 | bio_endio(bio, error); | |
1445 | return; | |
1446 | } | |
1447 | if (bio_empty_barrier(bio)) { | |
1448 | bio_endio(bio, 0); | |
1449 | return; | |
1450 | } | |
1451 | ||
1452 | __split_and_process_bio(md, bio); | |
1453 | ||
1454 | error = dm_flush(md); | |
1455 | ||
1456 | if (!error && md->barrier_error) | |
1457 | error = md->barrier_error; | |
1458 | ||
1459 | if (md->barrier_error != DM_ENDIO_REQUEUE) | |
1460 | bio_endio(bio, error); | |
1461 | } | |
1462 | ||
1da177e4 LT |
1463 | /* |
1464 | * Process the deferred bios | |
1465 | */ | |
ef208587 | 1466 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 1467 | { |
ef208587 MP |
1468 | struct mapped_device *md = container_of(work, struct mapped_device, |
1469 | work); | |
6d6f10df | 1470 | struct bio *c; |
1da177e4 | 1471 | |
ef208587 MP |
1472 | down_write(&md->io_lock); |
1473 | ||
3b00b203 | 1474 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
1475 | spin_lock_irq(&md->deferred_lock); |
1476 | c = bio_list_pop(&md->deferred); | |
1477 | spin_unlock_irq(&md->deferred_lock); | |
1478 | ||
1479 | if (!c) { | |
1eb787ec | 1480 | clear_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); |
df12ee99 AK |
1481 | break; |
1482 | } | |
022c2611 | 1483 | |
3b00b203 MP |
1484 | up_write(&md->io_lock); |
1485 | ||
af7e466a MP |
1486 | if (bio_barrier(c)) |
1487 | process_barrier(md, c); | |
1488 | else | |
1489 | __split_and_process_bio(md, c); | |
3b00b203 MP |
1490 | |
1491 | down_write(&md->io_lock); | |
022c2611 | 1492 | } |
73d410c0 | 1493 | |
ef208587 | 1494 | up_write(&md->io_lock); |
1da177e4 LT |
1495 | } |
1496 | ||
9a1fb464 | 1497 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 1498 | { |
3b00b203 MP |
1499 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1500 | smp_mb__after_clear_bit(); | |
53d5914f | 1501 | queue_work(md->wq, &md->work); |
304f3f6a MB |
1502 | } |
1503 | ||
1da177e4 LT |
1504 | /* |
1505 | * Swap in a new table (destroying old one). | |
1506 | */ | |
1507 | int dm_swap_table(struct mapped_device *md, struct dm_table *table) | |
1508 | { | |
93c534ae | 1509 | int r = -EINVAL; |
1da177e4 | 1510 | |
e61290a4 | 1511 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
1512 | |
1513 | /* device must be suspended */ | |
cf222b37 | 1514 | if (!dm_suspended(md)) |
93c534ae | 1515 | goto out; |
1da177e4 | 1516 | |
bfa152fa JN |
1517 | /* without bdev, the device size cannot be changed */ |
1518 | if (!md->suspended_bdev) | |
1519 | if (get_capacity(md->disk) != dm_table_get_size(table)) | |
1520 | goto out; | |
1521 | ||
1da177e4 LT |
1522 | __unbind(md); |
1523 | r = __bind(md, table); | |
1da177e4 | 1524 | |
93c534ae | 1525 | out: |
e61290a4 | 1526 | mutex_unlock(&md->suspend_lock); |
93c534ae | 1527 | return r; |
1da177e4 LT |
1528 | } |
1529 | ||
1530 | /* | |
1531 | * Functions to lock and unlock any filesystem running on the | |
1532 | * device. | |
1533 | */ | |
2ca3310e | 1534 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 1535 | { |
e39e2e95 | 1536 | int r; |
1da177e4 LT |
1537 | |
1538 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 1539 | |
e39e2e95 | 1540 | md->frozen_sb = freeze_bdev(md->suspended_bdev); |
dfbe03f6 | 1541 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 1542 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
1543 | md->frozen_sb = NULL; |
1544 | return r; | |
dfbe03f6 AK |
1545 | } |
1546 | ||
aa8d7c2f AK |
1547 | set_bit(DMF_FROZEN, &md->flags); |
1548 | ||
1da177e4 | 1549 | /* don't bdput right now, we don't want the bdev |
e39e2e95 | 1550 | * to go away while it is locked. |
1da177e4 LT |
1551 | */ |
1552 | return 0; | |
1553 | } | |
1554 | ||
2ca3310e | 1555 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 1556 | { |
aa8d7c2f AK |
1557 | if (!test_bit(DMF_FROZEN, &md->flags)) |
1558 | return; | |
1559 | ||
e39e2e95 | 1560 | thaw_bdev(md->suspended_bdev, md->frozen_sb); |
1da177e4 | 1561 | md->frozen_sb = NULL; |
aa8d7c2f | 1562 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
1563 | } |
1564 | ||
1565 | /* | |
1566 | * We need to be able to change a mapping table under a mounted | |
1567 | * filesystem. For example we might want to move some data in | |
1568 | * the background. Before the table can be swapped with | |
1569 | * dm_bind_table, dm_suspend must be called to flush any in | |
1570 | * flight bios and ensure that any further io gets deferred. | |
1571 | */ | |
a3d77d35 | 1572 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 1573 | { |
2ca3310e | 1574 | struct dm_table *map = NULL; |
46125c1c | 1575 | int r = 0; |
a3d77d35 | 1576 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 1577 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 1578 | |
e61290a4 | 1579 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1580 | |
73d410c0 MB |
1581 | if (dm_suspended(md)) { |
1582 | r = -EINVAL; | |
d287483d | 1583 | goto out_unlock; |
73d410c0 | 1584 | } |
1da177e4 LT |
1585 | |
1586 | map = dm_get_table(md); | |
1da177e4 | 1587 | |
2e93ccc1 KU |
1588 | /* |
1589 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
1590 | * This flag is cleared before dm_suspend returns. | |
1591 | */ | |
1592 | if (noflush) | |
1593 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
1594 | ||
cf222b37 AK |
1595 | /* This does not get reverted if there's an error later. */ |
1596 | dm_table_presuspend_targets(map); | |
1597 | ||
bfa152fa JN |
1598 | /* bdget() can stall if the pending I/Os are not flushed */ |
1599 | if (!noflush) { | |
1600 | md->suspended_bdev = bdget_disk(md->disk, 0); | |
1601 | if (!md->suspended_bdev) { | |
1602 | DMWARN("bdget failed in dm_suspend"); | |
1603 | r = -ENOMEM; | |
f431d966 | 1604 | goto out; |
bfa152fa | 1605 | } |
e39e2e95 | 1606 | |
6d6f10df MB |
1607 | /* |
1608 | * Flush I/O to the device. noflush supersedes do_lockfs, | |
1609 | * because lock_fs() needs to flush I/Os. | |
1610 | */ | |
1611 | if (do_lockfs) { | |
1612 | r = lock_fs(md); | |
1613 | if (r) | |
1614 | goto out; | |
1615 | } | |
aa8d7c2f | 1616 | } |
1da177e4 LT |
1617 | |
1618 | /* | |
3b00b203 MP |
1619 | * Here we must make sure that no processes are submitting requests |
1620 | * to target drivers i.e. no one may be executing | |
1621 | * __split_and_process_bio. This is called from dm_request and | |
1622 | * dm_wq_work. | |
1623 | * | |
1624 | * To get all processes out of __split_and_process_bio in dm_request, | |
1625 | * we take the write lock. To prevent any process from reentering | |
1626 | * __split_and_process_bio from dm_request, we set | |
1627 | * DMF_QUEUE_IO_TO_THREAD. | |
1628 | * | |
1629 | * To quiesce the thread (dm_wq_work), we set DMF_BLOCK_IO_FOR_SUSPEND | |
1630 | * and call flush_workqueue(md->wq). flush_workqueue will wait until | |
1631 | * dm_wq_work exits and DMF_BLOCK_IO_FOR_SUSPEND will prevent any | |
1632 | * further calls to __split_and_process_bio from dm_wq_work. | |
1da177e4 | 1633 | */ |
2ca3310e | 1634 | down_write(&md->io_lock); |
1eb787ec AK |
1635 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
1636 | set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags); | |
2ca3310e | 1637 | up_write(&md->io_lock); |
1da177e4 | 1638 | |
3b00b203 MP |
1639 | flush_workqueue(md->wq); |
1640 | ||
1da177e4 | 1641 | /* |
3b00b203 MP |
1642 | * At this point no more requests are entering target request routines. |
1643 | * We call dm_wait_for_completion to wait for all existing requests | |
1644 | * to finish. | |
1da177e4 | 1645 | */ |
401600df | 1646 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 1647 | |
2ca3310e | 1648 | down_write(&md->io_lock); |
6d6f10df | 1649 | if (noflush) |
022c2611 | 1650 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 1651 | up_write(&md->io_lock); |
2e93ccc1 | 1652 | |
1da177e4 | 1653 | /* were we interrupted ? */ |
46125c1c | 1654 | if (r < 0) { |
9a1fb464 | 1655 | dm_queue_flush(md); |
73d410c0 | 1656 | |
2ca3310e | 1657 | unlock_fs(md); |
2e93ccc1 | 1658 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 1659 | } |
1da177e4 | 1660 | |
3b00b203 MP |
1661 | /* |
1662 | * If dm_wait_for_completion returned 0, the device is completely | |
1663 | * quiescent now. There is no request-processing activity. All new | |
1664 | * requests are being added to md->deferred list. | |
1665 | */ | |
1666 | ||
cf222b37 | 1667 | dm_table_postsuspend_targets(map); |
1da177e4 | 1668 | |
2ca3310e | 1669 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 1670 | |
2ca3310e | 1671 | out: |
e39e2e95 AK |
1672 | if (r && md->suspended_bdev) { |
1673 | bdput(md->suspended_bdev); | |
1674 | md->suspended_bdev = NULL; | |
1675 | } | |
1676 | ||
2ca3310e | 1677 | dm_table_put(map); |
d287483d AK |
1678 | |
1679 | out_unlock: | |
e61290a4 | 1680 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 1681 | return r; |
1da177e4 LT |
1682 | } |
1683 | ||
1684 | int dm_resume(struct mapped_device *md) | |
1685 | { | |
cf222b37 | 1686 | int r = -EINVAL; |
cf222b37 | 1687 | struct dm_table *map = NULL; |
1da177e4 | 1688 | |
e61290a4 | 1689 | mutex_lock(&md->suspend_lock); |
2ca3310e | 1690 | if (!dm_suspended(md)) |
cf222b37 | 1691 | goto out; |
cf222b37 AK |
1692 | |
1693 | map = dm_get_table(md); | |
2ca3310e | 1694 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 1695 | goto out; |
1da177e4 | 1696 | |
8757b776 MB |
1697 | r = dm_table_resume_targets(map); |
1698 | if (r) | |
1699 | goto out; | |
2ca3310e | 1700 | |
9a1fb464 | 1701 | dm_queue_flush(md); |
2ca3310e AK |
1702 | |
1703 | unlock_fs(md); | |
1704 | ||
bfa152fa JN |
1705 | if (md->suspended_bdev) { |
1706 | bdput(md->suspended_bdev); | |
1707 | md->suspended_bdev = NULL; | |
1708 | } | |
e39e2e95 | 1709 | |
2ca3310e AK |
1710 | clear_bit(DMF_SUSPENDED, &md->flags); |
1711 | ||
1da177e4 | 1712 | dm_table_unplug_all(map); |
1da177e4 | 1713 | |
69267a30 | 1714 | dm_kobject_uevent(md); |
8560ed6f | 1715 | |
cf222b37 | 1716 | r = 0; |
2ca3310e | 1717 | |
cf222b37 AK |
1718 | out: |
1719 | dm_table_put(map); | |
e61290a4 | 1720 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 1721 | |
cf222b37 | 1722 | return r; |
1da177e4 LT |
1723 | } |
1724 | ||
1725 | /*----------------------------------------------------------------- | |
1726 | * Event notification. | |
1727 | *---------------------------------------------------------------*/ | |
69267a30 AK |
1728 | void dm_kobject_uevent(struct mapped_device *md) |
1729 | { | |
ed9e1982 | 1730 | kobject_uevent(&disk_to_dev(md->disk)->kobj, KOBJ_CHANGE); |
69267a30 AK |
1731 | } |
1732 | ||
7a8c3d3b MA |
1733 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
1734 | { | |
1735 | return atomic_add_return(1, &md->uevent_seq); | |
1736 | } | |
1737 | ||
1da177e4 LT |
1738 | uint32_t dm_get_event_nr(struct mapped_device *md) |
1739 | { | |
1740 | return atomic_read(&md->event_nr); | |
1741 | } | |
1742 | ||
1743 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
1744 | { | |
1745 | return wait_event_interruptible(md->eventq, | |
1746 | (event_nr != atomic_read(&md->event_nr))); | |
1747 | } | |
1748 | ||
7a8c3d3b MA |
1749 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
1750 | { | |
1751 | unsigned long flags; | |
1752 | ||
1753 | spin_lock_irqsave(&md->uevent_lock, flags); | |
1754 | list_add(elist, &md->uevent_list); | |
1755 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1756 | } | |
1757 | ||
1da177e4 LT |
1758 | /* |
1759 | * The gendisk is only valid as long as you have a reference | |
1760 | * count on 'md'. | |
1761 | */ | |
1762 | struct gendisk *dm_disk(struct mapped_device *md) | |
1763 | { | |
1764 | return md->disk; | |
1765 | } | |
1766 | ||
784aae73 MB |
1767 | struct kobject *dm_kobject(struct mapped_device *md) |
1768 | { | |
1769 | return &md->kobj; | |
1770 | } | |
1771 | ||
1772 | /* | |
1773 | * struct mapped_device should not be exported outside of dm.c | |
1774 | * so use this check to verify that kobj is part of md structure | |
1775 | */ | |
1776 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
1777 | { | |
1778 | struct mapped_device *md; | |
1779 | ||
1780 | md = container_of(kobj, struct mapped_device, kobj); | |
1781 | if (&md->kobj != kobj) | |
1782 | return NULL; | |
1783 | ||
1784 | dm_get(md); | |
1785 | return md; | |
1786 | } | |
1787 | ||
1da177e4 LT |
1788 | int dm_suspended(struct mapped_device *md) |
1789 | { | |
1790 | return test_bit(DMF_SUSPENDED, &md->flags); | |
1791 | } | |
1792 | ||
2e93ccc1 KU |
1793 | int dm_noflush_suspending(struct dm_target *ti) |
1794 | { | |
1795 | struct mapped_device *md = dm_table_get_md(ti->table); | |
1796 | int r = __noflush_suspending(md); | |
1797 | ||
1798 | dm_put(md); | |
1799 | ||
1800 | return r; | |
1801 | } | |
1802 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
1803 | ||
1da177e4 LT |
1804 | static struct block_device_operations dm_blk_dops = { |
1805 | .open = dm_blk_open, | |
1806 | .release = dm_blk_close, | |
aa129a22 | 1807 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 1808 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
1809 | .owner = THIS_MODULE |
1810 | }; | |
1811 | ||
1812 | EXPORT_SYMBOL(dm_get_mapinfo); | |
1813 | ||
1814 | /* | |
1815 | * module hooks | |
1816 | */ | |
1817 | module_init(dm_init); | |
1818 | module_exit(dm_exit); | |
1819 | ||
1820 | module_param(major, uint, 0); | |
1821 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
1822 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
1823 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
1824 | MODULE_LICENSE("GPL"); |