]>
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" | |
51e5b2bd | 9 | #include "dm-uevent.h" |
1da177e4 LT |
10 | |
11 | #include <linux/init.h> | |
12 | #include <linux/module.h> | |
48c9c27b | 13 | #include <linux/mutex.h> |
1da177e4 LT |
14 | #include <linux/moduleparam.h> |
15 | #include <linux/blkpg.h> | |
16 | #include <linux/bio.h> | |
17 | #include <linux/buffer_head.h> | |
18 | #include <linux/mempool.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/idr.h> | |
3ac51e74 | 21 | #include <linux/hdreg.h> |
3f77316d | 22 | #include <linux/delay.h> |
55782138 LZ |
23 | |
24 | #include <trace/events/block.h> | |
1da177e4 | 25 | |
72d94861 AK |
26 | #define DM_MSG_PREFIX "core" |
27 | ||
60935eb2 MB |
28 | /* |
29 | * Cookies are numeric values sent with CHANGE and REMOVE | |
30 | * uevents while resuming, removing or renaming the device. | |
31 | */ | |
32 | #define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE" | |
33 | #define DM_COOKIE_LENGTH 24 | |
34 | ||
1da177e4 LT |
35 | static const char *_name = DM_NAME; |
36 | ||
37 | static unsigned int major = 0; | |
38 | static unsigned int _major = 0; | |
39 | ||
d15b774c AK |
40 | static DEFINE_IDR(_minor_idr); |
41 | ||
f32c10b0 | 42 | static DEFINE_SPINLOCK(_minor_lock); |
1da177e4 | 43 | /* |
8fbf26ad | 44 | * For bio-based dm. |
1da177e4 LT |
45 | * One of these is allocated per bio. |
46 | */ | |
47 | struct dm_io { | |
48 | struct mapped_device *md; | |
49 | int error; | |
1da177e4 | 50 | atomic_t io_count; |
6ae2fa67 | 51 | struct bio *bio; |
3eaf840e | 52 | unsigned long start_time; |
f88fb981 | 53 | spinlock_t endio_lock; |
1da177e4 LT |
54 | }; |
55 | ||
56 | /* | |
8fbf26ad | 57 | * For bio-based dm. |
1da177e4 LT |
58 | * One of these is allocated per target within a bio. Hopefully |
59 | * this will be simplified out one day. | |
60 | */ | |
028867ac | 61 | struct dm_target_io { |
1da177e4 LT |
62 | struct dm_io *io; |
63 | struct dm_target *ti; | |
64 | union map_info info; | |
65 | }; | |
66 | ||
8fbf26ad KU |
67 | /* |
68 | * For request-based dm. | |
69 | * One of these is allocated per request. | |
70 | */ | |
71 | struct dm_rq_target_io { | |
72 | struct mapped_device *md; | |
73 | struct dm_target *ti; | |
74 | struct request *orig, clone; | |
75 | int error; | |
76 | union map_info info; | |
77 | }; | |
78 | ||
79 | /* | |
80 | * For request-based dm. | |
81 | * One of these is allocated per bio. | |
82 | */ | |
83 | struct dm_rq_clone_bio_info { | |
84 | struct bio *orig; | |
cec47e3d | 85 | struct dm_rq_target_io *tio; |
8fbf26ad KU |
86 | }; |
87 | ||
1da177e4 LT |
88 | union map_info *dm_get_mapinfo(struct bio *bio) |
89 | { | |
17b2f66f | 90 | if (bio && bio->bi_private) |
028867ac | 91 | return &((struct dm_target_io *)bio->bi_private)->info; |
17b2f66f | 92 | return NULL; |
1da177e4 LT |
93 | } |
94 | ||
cec47e3d KU |
95 | union map_info *dm_get_rq_mapinfo(struct request *rq) |
96 | { | |
97 | if (rq && rq->end_io_data) | |
98 | return &((struct dm_rq_target_io *)rq->end_io_data)->info; | |
99 | return NULL; | |
100 | } | |
101 | EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo); | |
102 | ||
ba61fdd1 JM |
103 | #define MINOR_ALLOCED ((void *)-1) |
104 | ||
1da177e4 LT |
105 | /* |
106 | * Bits for the md->flags field. | |
107 | */ | |
1eb787ec | 108 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
1da177e4 | 109 | #define DMF_SUSPENDED 1 |
aa8d7c2f | 110 | #define DMF_FROZEN 2 |
fba9f90e | 111 | #define DMF_FREEING 3 |
5c6bd75d | 112 | #define DMF_DELETING 4 |
2e93ccc1 | 113 | #define DMF_NOFLUSH_SUSPENDING 5 |
1da177e4 | 114 | |
304f3f6a MB |
115 | /* |
116 | * Work processed by per-device workqueue. | |
117 | */ | |
1da177e4 | 118 | struct mapped_device { |
2ca3310e | 119 | struct rw_semaphore io_lock; |
e61290a4 | 120 | struct mutex suspend_lock; |
1da177e4 LT |
121 | rwlock_t map_lock; |
122 | atomic_t holders; | |
5c6bd75d | 123 | atomic_t open_count; |
1da177e4 LT |
124 | |
125 | unsigned long flags; | |
126 | ||
165125e1 | 127 | struct request_queue *queue; |
a5664dad | 128 | unsigned type; |
4a0b4ddf | 129 | /* Protect queue and type against concurrent access. */ |
a5664dad MS |
130 | struct mutex type_lock; |
131 | ||
1da177e4 | 132 | struct gendisk *disk; |
7e51f257 | 133 | char name[16]; |
1da177e4 LT |
134 | |
135 | void *interface_ptr; | |
136 | ||
137 | /* | |
138 | * A list of ios that arrived while we were suspended. | |
139 | */ | |
316d315b | 140 | atomic_t pending[2]; |
1da177e4 | 141 | wait_queue_head_t wait; |
53d5914f | 142 | struct work_struct work; |
74859364 | 143 | struct bio_list deferred; |
022c2611 | 144 | spinlock_t deferred_lock; |
1da177e4 | 145 | |
af7e466a | 146 | /* |
29e4013d | 147 | * Processing queue (flush) |
304f3f6a MB |
148 | */ |
149 | struct workqueue_struct *wq; | |
150 | ||
1da177e4 LT |
151 | /* |
152 | * The current mapping. | |
153 | */ | |
154 | struct dm_table *map; | |
155 | ||
156 | /* | |
157 | * io objects are allocated from here. | |
158 | */ | |
159 | mempool_t *io_pool; | |
160 | mempool_t *tio_pool; | |
161 | ||
9faf400f SB |
162 | struct bio_set *bs; |
163 | ||
1da177e4 LT |
164 | /* |
165 | * Event handling. | |
166 | */ | |
167 | atomic_t event_nr; | |
168 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
169 | atomic_t uevent_seq; |
170 | struct list_head uevent_list; | |
171 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
172 | |
173 | /* | |
174 | * freeze/thaw support require holding onto a super block | |
175 | */ | |
176 | struct super_block *frozen_sb; | |
db8fef4f | 177 | struct block_device *bdev; |
3ac51e74 DW |
178 | |
179 | /* forced geometry settings */ | |
180 | struct hd_geometry geometry; | |
784aae73 | 181 | |
cec47e3d KU |
182 | /* For saving the address of __make_request for request based dm */ |
183 | make_request_fn *saved_make_request_fn; | |
184 | ||
784aae73 MB |
185 | /* sysfs handle */ |
186 | struct kobject kobj; | |
52b1fd5a | 187 | |
d87f4c14 TH |
188 | /* zero-length flush that will be cloned and submitted to targets */ |
189 | struct bio flush_bio; | |
1da177e4 LT |
190 | }; |
191 | ||
e6ee8c0b KU |
192 | /* |
193 | * For mempools pre-allocation at the table loading time. | |
194 | */ | |
195 | struct dm_md_mempools { | |
196 | mempool_t *io_pool; | |
197 | mempool_t *tio_pool; | |
198 | struct bio_set *bs; | |
199 | }; | |
200 | ||
1da177e4 | 201 | #define MIN_IOS 256 |
e18b890b CL |
202 | static struct kmem_cache *_io_cache; |
203 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
204 | static struct kmem_cache *_rq_tio_cache; |
205 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 206 | |
1da177e4 LT |
207 | static int __init local_init(void) |
208 | { | |
51157b4a | 209 | int r = -ENOMEM; |
1da177e4 | 210 | |
1da177e4 | 211 | /* allocate a slab for the dm_ios */ |
028867ac | 212 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 213 | if (!_io_cache) |
51157b4a | 214 | return r; |
1da177e4 LT |
215 | |
216 | /* allocate a slab for the target ios */ | |
028867ac | 217 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
218 | if (!_tio_cache) |
219 | goto out_free_io_cache; | |
1da177e4 | 220 | |
8fbf26ad KU |
221 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
222 | if (!_rq_tio_cache) | |
223 | goto out_free_tio_cache; | |
224 | ||
225 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
226 | if (!_rq_bio_info_cache) | |
227 | goto out_free_rq_tio_cache; | |
228 | ||
51e5b2bd | 229 | r = dm_uevent_init(); |
51157b4a | 230 | if (r) |
8fbf26ad | 231 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 232 | |
1da177e4 LT |
233 | _major = major; |
234 | r = register_blkdev(_major, _name); | |
51157b4a KU |
235 | if (r < 0) |
236 | goto out_uevent_exit; | |
1da177e4 LT |
237 | |
238 | if (!_major) | |
239 | _major = r; | |
240 | ||
241 | return 0; | |
51157b4a KU |
242 | |
243 | out_uevent_exit: | |
244 | dm_uevent_exit(); | |
8fbf26ad KU |
245 | out_free_rq_bio_info_cache: |
246 | kmem_cache_destroy(_rq_bio_info_cache); | |
247 | out_free_rq_tio_cache: | |
248 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
249 | out_free_tio_cache: |
250 | kmem_cache_destroy(_tio_cache); | |
251 | out_free_io_cache: | |
252 | kmem_cache_destroy(_io_cache); | |
253 | ||
254 | return r; | |
1da177e4 LT |
255 | } |
256 | ||
257 | static void local_exit(void) | |
258 | { | |
8fbf26ad KU |
259 | kmem_cache_destroy(_rq_bio_info_cache); |
260 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
261 | kmem_cache_destroy(_tio_cache); |
262 | kmem_cache_destroy(_io_cache); | |
00d59405 | 263 | unregister_blkdev(_major, _name); |
51e5b2bd | 264 | dm_uevent_exit(); |
1da177e4 LT |
265 | |
266 | _major = 0; | |
267 | ||
268 | DMINFO("cleaned up"); | |
269 | } | |
270 | ||
b9249e55 | 271 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
272 | local_init, |
273 | dm_target_init, | |
274 | dm_linear_init, | |
275 | dm_stripe_init, | |
952b3557 | 276 | dm_io_init, |
945fa4d2 | 277 | dm_kcopyd_init, |
1da177e4 LT |
278 | dm_interface_init, |
279 | }; | |
280 | ||
b9249e55 | 281 | static void (*_exits[])(void) = { |
1da177e4 LT |
282 | local_exit, |
283 | dm_target_exit, | |
284 | dm_linear_exit, | |
285 | dm_stripe_exit, | |
952b3557 | 286 | dm_io_exit, |
945fa4d2 | 287 | dm_kcopyd_exit, |
1da177e4 LT |
288 | dm_interface_exit, |
289 | }; | |
290 | ||
291 | static int __init dm_init(void) | |
292 | { | |
293 | const int count = ARRAY_SIZE(_inits); | |
294 | ||
295 | int r, i; | |
296 | ||
297 | for (i = 0; i < count; i++) { | |
298 | r = _inits[i](); | |
299 | if (r) | |
300 | goto bad; | |
301 | } | |
302 | ||
303 | return 0; | |
304 | ||
305 | bad: | |
306 | while (i--) | |
307 | _exits[i](); | |
308 | ||
309 | return r; | |
310 | } | |
311 | ||
312 | static void __exit dm_exit(void) | |
313 | { | |
314 | int i = ARRAY_SIZE(_exits); | |
315 | ||
316 | while (i--) | |
317 | _exits[i](); | |
d15b774c AK |
318 | |
319 | /* | |
320 | * Should be empty by this point. | |
321 | */ | |
322 | idr_remove_all(&_minor_idr); | |
323 | idr_destroy(&_minor_idr); | |
1da177e4 LT |
324 | } |
325 | ||
326 | /* | |
327 | * Block device functions | |
328 | */ | |
432a212c MA |
329 | int dm_deleting_md(struct mapped_device *md) |
330 | { | |
331 | return test_bit(DMF_DELETING, &md->flags); | |
332 | } | |
333 | ||
fe5f9f2c | 334 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
335 | { |
336 | struct mapped_device *md; | |
337 | ||
fba9f90e JM |
338 | spin_lock(&_minor_lock); |
339 | ||
fe5f9f2c | 340 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
341 | if (!md) |
342 | goto out; | |
343 | ||
5c6bd75d | 344 | if (test_bit(DMF_FREEING, &md->flags) || |
432a212c | 345 | dm_deleting_md(md)) { |
fba9f90e JM |
346 | md = NULL; |
347 | goto out; | |
348 | } | |
349 | ||
1da177e4 | 350 | dm_get(md); |
5c6bd75d | 351 | atomic_inc(&md->open_count); |
fba9f90e JM |
352 | |
353 | out: | |
354 | spin_unlock(&_minor_lock); | |
355 | ||
356 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
357 | } |
358 | ||
fe5f9f2c | 359 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 360 | { |
fe5f9f2c | 361 | struct mapped_device *md = disk->private_data; |
6e9624b8 | 362 | |
4a1aeb98 MB |
363 | spin_lock(&_minor_lock); |
364 | ||
5c6bd75d | 365 | atomic_dec(&md->open_count); |
1da177e4 | 366 | dm_put(md); |
4a1aeb98 MB |
367 | |
368 | spin_unlock(&_minor_lock); | |
6e9624b8 | 369 | |
1da177e4 LT |
370 | return 0; |
371 | } | |
372 | ||
5c6bd75d AK |
373 | int dm_open_count(struct mapped_device *md) |
374 | { | |
375 | return atomic_read(&md->open_count); | |
376 | } | |
377 | ||
378 | /* | |
379 | * Guarantees nothing is using the device before it's deleted. | |
380 | */ | |
381 | int dm_lock_for_deletion(struct mapped_device *md) | |
382 | { | |
383 | int r = 0; | |
384 | ||
385 | spin_lock(&_minor_lock); | |
386 | ||
387 | if (dm_open_count(md)) | |
388 | r = -EBUSY; | |
389 | else | |
390 | set_bit(DMF_DELETING, &md->flags); | |
391 | ||
392 | spin_unlock(&_minor_lock); | |
393 | ||
394 | return r; | |
395 | } | |
396 | ||
3ac51e74 DW |
397 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
398 | { | |
399 | struct mapped_device *md = bdev->bd_disk->private_data; | |
400 | ||
401 | return dm_get_geometry(md, geo); | |
402 | } | |
403 | ||
fe5f9f2c | 404 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
405 | unsigned int cmd, unsigned long arg) |
406 | { | |
fe5f9f2c | 407 | struct mapped_device *md = bdev->bd_disk->private_data; |
7c666411 | 408 | struct dm_table *map = dm_get_live_table(md); |
aa129a22 MB |
409 | struct dm_target *tgt; |
410 | int r = -ENOTTY; | |
411 | ||
aa129a22 MB |
412 | if (!map || !dm_table_get_size(map)) |
413 | goto out; | |
414 | ||
415 | /* We only support devices that have a single target */ | |
416 | if (dm_table_get_num_targets(map) != 1) | |
417 | goto out; | |
418 | ||
419 | tgt = dm_table_get_target(map, 0); | |
420 | ||
4f186f8b | 421 | if (dm_suspended_md(md)) { |
aa129a22 MB |
422 | r = -EAGAIN; |
423 | goto out; | |
424 | } | |
425 | ||
426 | if (tgt->type->ioctl) | |
647b3d00 | 427 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
428 | |
429 | out: | |
430 | dm_table_put(map); | |
431 | ||
aa129a22 MB |
432 | return r; |
433 | } | |
434 | ||
028867ac | 435 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
436 | { |
437 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
438 | } | |
439 | ||
028867ac | 440 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
441 | { |
442 | mempool_free(io, md->io_pool); | |
443 | } | |
444 | ||
028867ac | 445 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
446 | { |
447 | mempool_free(tio, md->tio_pool); | |
448 | } | |
449 | ||
08885643 KU |
450 | static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md, |
451 | gfp_t gfp_mask) | |
cec47e3d | 452 | { |
08885643 | 453 | return mempool_alloc(md->tio_pool, gfp_mask); |
cec47e3d KU |
454 | } |
455 | ||
456 | static void free_rq_tio(struct dm_rq_target_io *tio) | |
457 | { | |
458 | mempool_free(tio, tio->md->tio_pool); | |
459 | } | |
460 | ||
461 | static struct dm_rq_clone_bio_info *alloc_bio_info(struct mapped_device *md) | |
462 | { | |
463 | return mempool_alloc(md->io_pool, GFP_ATOMIC); | |
464 | } | |
465 | ||
466 | static void free_bio_info(struct dm_rq_clone_bio_info *info) | |
467 | { | |
468 | mempool_free(info, info->tio->md->io_pool); | |
469 | } | |
470 | ||
90abb8c4 KU |
471 | static int md_in_flight(struct mapped_device *md) |
472 | { | |
473 | return atomic_read(&md->pending[READ]) + | |
474 | atomic_read(&md->pending[WRITE]); | |
475 | } | |
476 | ||
3eaf840e JNN |
477 | static void start_io_acct(struct dm_io *io) |
478 | { | |
479 | struct mapped_device *md = io->md; | |
c9959059 | 480 | int cpu; |
316d315b | 481 | int rw = bio_data_dir(io->bio); |
3eaf840e JNN |
482 | |
483 | io->start_time = jiffies; | |
484 | ||
074a7aca TH |
485 | cpu = part_stat_lock(); |
486 | part_round_stats(cpu, &dm_disk(md)->part0); | |
487 | part_stat_unlock(); | |
1e9bb880 SL |
488 | atomic_set(&dm_disk(md)->part0.in_flight[rw], |
489 | atomic_inc_return(&md->pending[rw])); | |
3eaf840e JNN |
490 | } |
491 | ||
d221d2e7 | 492 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
493 | { |
494 | struct mapped_device *md = io->md; | |
495 | struct bio *bio = io->bio; | |
496 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 497 | int pending, cpu; |
3eaf840e JNN |
498 | int rw = bio_data_dir(bio); |
499 | ||
074a7aca TH |
500 | cpu = part_stat_lock(); |
501 | part_round_stats(cpu, &dm_disk(md)->part0); | |
502 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
503 | part_stat_unlock(); | |
3eaf840e | 504 | |
af7e466a MP |
505 | /* |
506 | * After this is decremented the bio must not be touched if it is | |
d87f4c14 | 507 | * a flush. |
af7e466a | 508 | */ |
1e9bb880 SL |
509 | pending = atomic_dec_return(&md->pending[rw]); |
510 | atomic_set(&dm_disk(md)->part0.in_flight[rw], pending); | |
316d315b | 511 | pending += atomic_read(&md->pending[rw^0x1]); |
3eaf840e | 512 | |
d221d2e7 MP |
513 | /* nudge anyone waiting on suspend queue */ |
514 | if (!pending) | |
515 | wake_up(&md->wait); | |
3eaf840e JNN |
516 | } |
517 | ||
1da177e4 LT |
518 | /* |
519 | * Add the bio to the list of deferred io. | |
520 | */ | |
92c63902 | 521 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 522 | { |
05447420 | 523 | unsigned long flags; |
1da177e4 | 524 | |
05447420 | 525 | spin_lock_irqsave(&md->deferred_lock, flags); |
1da177e4 | 526 | bio_list_add(&md->deferred, bio); |
05447420 | 527 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
6a8736d1 | 528 | queue_work(md->wq, &md->work); |
1da177e4 LT |
529 | } |
530 | ||
531 | /* | |
532 | * Everyone (including functions in this file), should use this | |
533 | * function to access the md->map field, and make sure they call | |
534 | * dm_table_put() when finished. | |
535 | */ | |
7c666411 | 536 | struct dm_table *dm_get_live_table(struct mapped_device *md) |
1da177e4 LT |
537 | { |
538 | struct dm_table *t; | |
523d9297 | 539 | unsigned long flags; |
1da177e4 | 540 | |
523d9297 | 541 | read_lock_irqsave(&md->map_lock, flags); |
1da177e4 LT |
542 | t = md->map; |
543 | if (t) | |
544 | dm_table_get(t); | |
523d9297 | 545 | read_unlock_irqrestore(&md->map_lock, flags); |
1da177e4 LT |
546 | |
547 | return t; | |
548 | } | |
549 | ||
3ac51e74 DW |
550 | /* |
551 | * Get the geometry associated with a dm device | |
552 | */ | |
553 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
554 | { | |
555 | *geo = md->geometry; | |
556 | ||
557 | return 0; | |
558 | } | |
559 | ||
560 | /* | |
561 | * Set the geometry of a device. | |
562 | */ | |
563 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
564 | { | |
565 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
566 | ||
567 | if (geo->start > sz) { | |
568 | DMWARN("Start sector is beyond the geometry limits."); | |
569 | return -EINVAL; | |
570 | } | |
571 | ||
572 | md->geometry = *geo; | |
573 | ||
574 | return 0; | |
575 | } | |
576 | ||
1da177e4 LT |
577 | /*----------------------------------------------------------------- |
578 | * CRUD START: | |
579 | * A more elegant soln is in the works that uses the queue | |
580 | * merge fn, unfortunately there are a couple of changes to | |
581 | * the block layer that I want to make for this. So in the | |
582 | * interests of getting something for people to use I give | |
583 | * you this clearly demarcated crap. | |
584 | *---------------------------------------------------------------*/ | |
585 | ||
2e93ccc1 KU |
586 | static int __noflush_suspending(struct mapped_device *md) |
587 | { | |
588 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
589 | } | |
590 | ||
1da177e4 LT |
591 | /* |
592 | * Decrements the number of outstanding ios that a bio has been | |
593 | * cloned into, completing the original io if necc. | |
594 | */ | |
858119e1 | 595 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 596 | { |
2e93ccc1 | 597 | unsigned long flags; |
b35f8caa MB |
598 | int io_error; |
599 | struct bio *bio; | |
600 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
601 | |
602 | /* Push-back supersedes any I/O errors */ | |
f88fb981 KU |
603 | if (unlikely(error)) { |
604 | spin_lock_irqsave(&io->endio_lock, flags); | |
605 | if (!(io->error > 0 && __noflush_suspending(md))) | |
606 | io->error = error; | |
607 | spin_unlock_irqrestore(&io->endio_lock, flags); | |
608 | } | |
1da177e4 LT |
609 | |
610 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
611 | if (io->error == DM_ENDIO_REQUEUE) { |
612 | /* | |
613 | * Target requested pushing back the I/O. | |
2e93ccc1 | 614 | */ |
022c2611 | 615 | spin_lock_irqsave(&md->deferred_lock, flags); |
6a8736d1 TH |
616 | if (__noflush_suspending(md)) |
617 | bio_list_add_head(&md->deferred, io->bio); | |
618 | else | |
2e93ccc1 KU |
619 | /* noflush suspend was interrupted. */ |
620 | io->error = -EIO; | |
022c2611 | 621 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
622 | } |
623 | ||
b35f8caa MB |
624 | io_error = io->error; |
625 | bio = io->bio; | |
6a8736d1 TH |
626 | end_io_acct(io); |
627 | free_io(md, io); | |
628 | ||
629 | if (io_error == DM_ENDIO_REQUEUE) | |
630 | return; | |
2e93ccc1 | 631 | |
b372d360 | 632 | if ((bio->bi_rw & REQ_FLUSH) && bio->bi_size) { |
af7e466a | 633 | /* |
6a8736d1 TH |
634 | * Preflush done for flush with data, reissue |
635 | * without REQ_FLUSH. | |
af7e466a | 636 | */ |
6a8736d1 TH |
637 | bio->bi_rw &= ~REQ_FLUSH; |
638 | queue_io(md, bio); | |
af7e466a | 639 | } else { |
b372d360 | 640 | /* done with normal IO or empty flush */ |
b7908c10 | 641 | trace_block_bio_complete(md->queue, bio, io_error); |
b372d360 | 642 | bio_endio(bio, io_error); |
b35f8caa | 643 | } |
1da177e4 LT |
644 | } |
645 | } | |
646 | ||
6712ecf8 | 647 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
648 | { |
649 | int r = 0; | |
028867ac | 650 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 651 | struct dm_io *io = tio->io; |
9faf400f | 652 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
653 | dm_endio_fn endio = tio->ti->type->end_io; |
654 | ||
1da177e4 LT |
655 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
656 | error = -EIO; | |
657 | ||
658 | if (endio) { | |
659 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
660 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
661 | /* | |
662 | * error and requeue request are handled | |
663 | * in dec_pending(). | |
664 | */ | |
1da177e4 | 665 | error = r; |
45cbcd79 KU |
666 | else if (r == DM_ENDIO_INCOMPLETE) |
667 | /* The target will handle the io */ | |
6712ecf8 | 668 | return; |
45cbcd79 KU |
669 | else if (r) { |
670 | DMWARN("unimplemented target endio return value: %d", r); | |
671 | BUG(); | |
672 | } | |
1da177e4 LT |
673 | } |
674 | ||
9faf400f SB |
675 | /* |
676 | * Store md for cleanup instead of tio which is about to get freed. | |
677 | */ | |
678 | bio->bi_private = md->bs; | |
679 | ||
9faf400f | 680 | free_tio(md, tio); |
b35f8caa MB |
681 | bio_put(bio); |
682 | dec_pending(io, error); | |
1da177e4 LT |
683 | } |
684 | ||
cec47e3d KU |
685 | /* |
686 | * Partial completion handling for request-based dm | |
687 | */ | |
688 | static void end_clone_bio(struct bio *clone, int error) | |
689 | { | |
690 | struct dm_rq_clone_bio_info *info = clone->bi_private; | |
691 | struct dm_rq_target_io *tio = info->tio; | |
692 | struct bio *bio = info->orig; | |
693 | unsigned int nr_bytes = info->orig->bi_size; | |
694 | ||
695 | bio_put(clone); | |
696 | ||
697 | if (tio->error) | |
698 | /* | |
699 | * An error has already been detected on the request. | |
700 | * Once error occurred, just let clone->end_io() handle | |
701 | * the remainder. | |
702 | */ | |
703 | return; | |
704 | else if (error) { | |
705 | /* | |
706 | * Don't notice the error to the upper layer yet. | |
707 | * The error handling decision is made by the target driver, | |
708 | * when the request is completed. | |
709 | */ | |
710 | tio->error = error; | |
711 | return; | |
712 | } | |
713 | ||
714 | /* | |
715 | * I/O for the bio successfully completed. | |
716 | * Notice the data completion to the upper layer. | |
717 | */ | |
718 | ||
719 | /* | |
720 | * bios are processed from the head of the list. | |
721 | * So the completing bio should always be rq->bio. | |
722 | * If it's not, something wrong is happening. | |
723 | */ | |
724 | if (tio->orig->bio != bio) | |
725 | DMERR("bio completion is going in the middle of the request"); | |
726 | ||
727 | /* | |
728 | * Update the original request. | |
729 | * Do not use blk_end_request() here, because it may complete | |
730 | * the original request before the clone, and break the ordering. | |
731 | */ | |
732 | blk_update_request(tio->orig, 0, nr_bytes); | |
733 | } | |
734 | ||
735 | /* | |
736 | * Don't touch any member of the md after calling this function because | |
737 | * the md may be freed in dm_put() at the end of this function. | |
738 | * Or do dm_get() before calling this function and dm_put() later. | |
739 | */ | |
b4324fee | 740 | static void rq_completed(struct mapped_device *md, int rw, int run_queue) |
cec47e3d | 741 | { |
b4324fee | 742 | atomic_dec(&md->pending[rw]); |
cec47e3d KU |
743 | |
744 | /* nudge anyone waiting on suspend queue */ | |
b4324fee | 745 | if (!md_in_flight(md)) |
cec47e3d KU |
746 | wake_up(&md->wait); |
747 | ||
748 | if (run_queue) | |
b4324fee | 749 | blk_run_queue(md->queue); |
cec47e3d KU |
750 | |
751 | /* | |
752 | * dm_put() must be at the end of this function. See the comment above | |
753 | */ | |
754 | dm_put(md); | |
755 | } | |
756 | ||
a77e28c7 KU |
757 | static void free_rq_clone(struct request *clone) |
758 | { | |
759 | struct dm_rq_target_io *tio = clone->end_io_data; | |
760 | ||
761 | blk_rq_unprep_clone(clone); | |
762 | free_rq_tio(tio); | |
763 | } | |
764 | ||
980691e5 KU |
765 | /* |
766 | * Complete the clone and the original request. | |
767 | * Must be called without queue lock. | |
768 | */ | |
769 | static void dm_end_request(struct request *clone, int error) | |
770 | { | |
771 | int rw = rq_data_dir(clone); | |
772 | struct dm_rq_target_io *tio = clone->end_io_data; | |
773 | struct mapped_device *md = tio->md; | |
774 | struct request *rq = tio->orig; | |
775 | ||
29e4013d | 776 | if (rq->cmd_type == REQ_TYPE_BLOCK_PC) { |
980691e5 KU |
777 | rq->errors = clone->errors; |
778 | rq->resid_len = clone->resid_len; | |
779 | ||
780 | if (rq->sense) | |
781 | /* | |
782 | * We are using the sense buffer of the original | |
783 | * request. | |
784 | * So setting the length of the sense data is enough. | |
785 | */ | |
786 | rq->sense_len = clone->sense_len; | |
787 | } | |
788 | ||
789 | free_rq_clone(clone); | |
29e4013d TH |
790 | blk_end_request_all(rq, error); |
791 | rq_completed(md, rw, true); | |
980691e5 KU |
792 | } |
793 | ||
cec47e3d KU |
794 | static void dm_unprep_request(struct request *rq) |
795 | { | |
796 | struct request *clone = rq->special; | |
cec47e3d KU |
797 | |
798 | rq->special = NULL; | |
799 | rq->cmd_flags &= ~REQ_DONTPREP; | |
800 | ||
a77e28c7 | 801 | free_rq_clone(clone); |
cec47e3d KU |
802 | } |
803 | ||
804 | /* | |
805 | * Requeue the original request of a clone. | |
806 | */ | |
807 | void dm_requeue_unmapped_request(struct request *clone) | |
808 | { | |
b4324fee | 809 | int rw = rq_data_dir(clone); |
cec47e3d KU |
810 | struct dm_rq_target_io *tio = clone->end_io_data; |
811 | struct mapped_device *md = tio->md; | |
812 | struct request *rq = tio->orig; | |
813 | struct request_queue *q = rq->q; | |
814 | unsigned long flags; | |
815 | ||
816 | dm_unprep_request(rq); | |
817 | ||
818 | spin_lock_irqsave(q->queue_lock, flags); | |
cec47e3d KU |
819 | blk_requeue_request(q, rq); |
820 | spin_unlock_irqrestore(q->queue_lock, flags); | |
821 | ||
b4324fee | 822 | rq_completed(md, rw, 0); |
cec47e3d KU |
823 | } |
824 | EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request); | |
825 | ||
826 | static void __stop_queue(struct request_queue *q) | |
827 | { | |
828 | blk_stop_queue(q); | |
829 | } | |
830 | ||
831 | static void stop_queue(struct request_queue *q) | |
832 | { | |
833 | unsigned long flags; | |
834 | ||
835 | spin_lock_irqsave(q->queue_lock, flags); | |
836 | __stop_queue(q); | |
837 | spin_unlock_irqrestore(q->queue_lock, flags); | |
838 | } | |
839 | ||
840 | static void __start_queue(struct request_queue *q) | |
841 | { | |
842 | if (blk_queue_stopped(q)) | |
843 | blk_start_queue(q); | |
844 | } | |
845 | ||
846 | static void start_queue(struct request_queue *q) | |
847 | { | |
848 | unsigned long flags; | |
849 | ||
850 | spin_lock_irqsave(q->queue_lock, flags); | |
851 | __start_queue(q); | |
852 | spin_unlock_irqrestore(q->queue_lock, flags); | |
853 | } | |
854 | ||
11a68244 | 855 | static void dm_done(struct request *clone, int error, bool mapped) |
cec47e3d | 856 | { |
11a68244 | 857 | int r = error; |
cec47e3d KU |
858 | struct dm_rq_target_io *tio = clone->end_io_data; |
859 | dm_request_endio_fn rq_end_io = tio->ti->type->rq_end_io; | |
cec47e3d | 860 | |
11a68244 KU |
861 | if (mapped && rq_end_io) |
862 | r = rq_end_io(tio->ti, clone, error, &tio->info); | |
cec47e3d | 863 | |
11a68244 | 864 | if (r <= 0) |
cec47e3d | 865 | /* The target wants to complete the I/O */ |
11a68244 KU |
866 | dm_end_request(clone, r); |
867 | else if (r == DM_ENDIO_INCOMPLETE) | |
cec47e3d KU |
868 | /* The target will handle the I/O */ |
869 | return; | |
11a68244 | 870 | else if (r == DM_ENDIO_REQUEUE) |
cec47e3d KU |
871 | /* The target wants to requeue the I/O */ |
872 | dm_requeue_unmapped_request(clone); | |
873 | else { | |
11a68244 | 874 | DMWARN("unimplemented target endio return value: %d", r); |
cec47e3d KU |
875 | BUG(); |
876 | } | |
877 | } | |
878 | ||
11a68244 KU |
879 | /* |
880 | * Request completion handler for request-based dm | |
881 | */ | |
882 | static void dm_softirq_done(struct request *rq) | |
883 | { | |
884 | bool mapped = true; | |
885 | struct request *clone = rq->completion_data; | |
886 | struct dm_rq_target_io *tio = clone->end_io_data; | |
887 | ||
888 | if (rq->cmd_flags & REQ_FAILED) | |
889 | mapped = false; | |
890 | ||
891 | dm_done(clone, tio->error, mapped); | |
892 | } | |
893 | ||
cec47e3d KU |
894 | /* |
895 | * Complete the clone and the original request with the error status | |
896 | * through softirq context. | |
897 | */ | |
898 | static void dm_complete_request(struct request *clone, int error) | |
899 | { | |
900 | struct dm_rq_target_io *tio = clone->end_io_data; | |
901 | struct request *rq = tio->orig; | |
902 | ||
903 | tio->error = error; | |
904 | rq->completion_data = clone; | |
905 | blk_complete_request(rq); | |
906 | } | |
907 | ||
908 | /* | |
909 | * Complete the not-mapped clone and the original request with the error status | |
910 | * through softirq context. | |
911 | * Target's rq_end_io() function isn't called. | |
912 | * This may be used when the target's map_rq() function fails. | |
913 | */ | |
914 | void dm_kill_unmapped_request(struct request *clone, int error) | |
915 | { | |
916 | struct dm_rq_target_io *tio = clone->end_io_data; | |
917 | struct request *rq = tio->orig; | |
918 | ||
919 | rq->cmd_flags |= REQ_FAILED; | |
920 | dm_complete_request(clone, error); | |
921 | } | |
922 | EXPORT_SYMBOL_GPL(dm_kill_unmapped_request); | |
923 | ||
924 | /* | |
925 | * Called with the queue lock held | |
926 | */ | |
927 | static void end_clone_request(struct request *clone, int error) | |
928 | { | |
929 | /* | |
930 | * For just cleaning up the information of the queue in which | |
931 | * the clone was dispatched. | |
932 | * The clone is *NOT* freed actually here because it is alloced from | |
933 | * dm own mempool and REQ_ALLOCED isn't set in clone->cmd_flags. | |
934 | */ | |
935 | __blk_put_request(clone->q, clone); | |
936 | ||
937 | /* | |
938 | * Actual request completion is done in a softirq context which doesn't | |
939 | * hold the queue lock. Otherwise, deadlock could occur because: | |
940 | * - another request may be submitted by the upper level driver | |
941 | * of the stacking during the completion | |
942 | * - the submission which requires queue lock may be done | |
943 | * against this queue | |
944 | */ | |
945 | dm_complete_request(clone, error); | |
946 | } | |
947 | ||
56a67df7 MS |
948 | /* |
949 | * Return maximum size of I/O possible at the supplied sector up to the current | |
950 | * target boundary. | |
951 | */ | |
952 | static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti) | |
953 | { | |
954 | sector_t target_offset = dm_target_offset(ti, sector); | |
955 | ||
956 | return ti->len - target_offset; | |
957 | } | |
958 | ||
959 | static sector_t max_io_len(sector_t sector, struct dm_target *ti) | |
1da177e4 | 960 | { |
56a67df7 | 961 | sector_t len = max_io_len_target_boundary(sector, ti); |
1da177e4 LT |
962 | |
963 | /* | |
964 | * Does the target need to split even further ? | |
965 | */ | |
966 | if (ti->split_io) { | |
967 | sector_t boundary; | |
56a67df7 | 968 | sector_t offset = dm_target_offset(ti, sector); |
1da177e4 LT |
969 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) |
970 | - offset; | |
971 | if (len > boundary) | |
972 | len = boundary; | |
973 | } | |
974 | ||
975 | return len; | |
976 | } | |
977 | ||
978 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 979 | struct dm_target_io *tio) |
1da177e4 LT |
980 | { |
981 | int r; | |
2056a782 | 982 | sector_t sector; |
9faf400f | 983 | struct mapped_device *md; |
1da177e4 | 984 | |
1da177e4 LT |
985 | clone->bi_end_io = clone_endio; |
986 | clone->bi_private = tio; | |
987 | ||
988 | /* | |
989 | * Map the clone. If r == 0 we don't need to do | |
990 | * anything, the target has assumed ownership of | |
991 | * this io. | |
992 | */ | |
993 | atomic_inc(&tio->io->io_count); | |
2056a782 | 994 | sector = clone->bi_sector; |
1da177e4 | 995 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 996 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 997 | /* the bio has been remapped so dispatch it */ |
2056a782 | 998 | |
d07335e5 MS |
999 | trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone, |
1000 | tio->io->bio->bi_bdev->bd_dev, sector); | |
2056a782 | 1001 | |
1da177e4 | 1002 | generic_make_request(clone); |
2e93ccc1 KU |
1003 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
1004 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
1005 | md = tio->io->md; |
1006 | dec_pending(tio->io, r); | |
1007 | /* | |
1008 | * Store bio_set for cleanup. | |
1009 | */ | |
1010 | clone->bi_private = md->bs; | |
1da177e4 | 1011 | bio_put(clone); |
9faf400f | 1012 | free_tio(md, tio); |
45cbcd79 KU |
1013 | } else if (r) { |
1014 | DMWARN("unimplemented target map return value: %d", r); | |
1015 | BUG(); | |
1da177e4 LT |
1016 | } |
1017 | } | |
1018 | ||
1019 | struct clone_info { | |
1020 | struct mapped_device *md; | |
1021 | struct dm_table *map; | |
1022 | struct bio *bio; | |
1023 | struct dm_io *io; | |
1024 | sector_t sector; | |
1025 | sector_t sector_count; | |
1026 | unsigned short idx; | |
1027 | }; | |
1028 | ||
3676347a PO |
1029 | static void dm_bio_destructor(struct bio *bio) |
1030 | { | |
9faf400f SB |
1031 | struct bio_set *bs = bio->bi_private; |
1032 | ||
1033 | bio_free(bio, bs); | |
3676347a PO |
1034 | } |
1035 | ||
1da177e4 | 1036 | /* |
d87f4c14 | 1037 | * Creates a little bio that just does part of a bvec. |
1da177e4 LT |
1038 | */ |
1039 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
1040 | unsigned short idx, unsigned int offset, | |
9faf400f | 1041 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
1042 | { |
1043 | struct bio *clone; | |
1044 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
1045 | ||
9faf400f | 1046 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 1047 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
1048 | *clone->bi_io_vec = *bv; |
1049 | ||
1050 | clone->bi_sector = sector; | |
1051 | clone->bi_bdev = bio->bi_bdev; | |
d87f4c14 | 1052 | clone->bi_rw = bio->bi_rw; |
1da177e4 LT |
1053 | clone->bi_vcnt = 1; |
1054 | clone->bi_size = to_bytes(len); | |
1055 | clone->bi_io_vec->bv_offset = offset; | |
1056 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 1057 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 1058 | |
9c47008d | 1059 | if (bio_integrity(bio)) { |
7878cba9 | 1060 | bio_integrity_clone(clone, bio, GFP_NOIO, bs); |
9c47008d MP |
1061 | bio_integrity_trim(clone, |
1062 | bio_sector_offset(bio, idx, offset), len); | |
1063 | } | |
1064 | ||
1da177e4 LT |
1065 | return clone; |
1066 | } | |
1067 | ||
1068 | /* | |
1069 | * Creates a bio that consists of range of complete bvecs. | |
1070 | */ | |
1071 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
1072 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 1073 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
1074 | { |
1075 | struct bio *clone; | |
1076 | ||
9faf400f SB |
1077 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
1078 | __bio_clone(clone, bio); | |
1079 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
1080 | clone->bi_sector = sector; |
1081 | clone->bi_idx = idx; | |
1082 | clone->bi_vcnt = idx + bv_count; | |
1083 | clone->bi_size = to_bytes(len); | |
1084 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
1085 | ||
9c47008d | 1086 | if (bio_integrity(bio)) { |
7878cba9 | 1087 | bio_integrity_clone(clone, bio, GFP_NOIO, bs); |
9c47008d MP |
1088 | |
1089 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
1090 | bio_integrity_trim(clone, | |
1091 | bio_sector_offset(bio, idx, 0), len); | |
1092 | } | |
1093 | ||
1da177e4 LT |
1094 | return clone; |
1095 | } | |
1096 | ||
9015df24 AK |
1097 | static struct dm_target_io *alloc_tio(struct clone_info *ci, |
1098 | struct dm_target *ti) | |
f9ab94ce | 1099 | { |
9015df24 | 1100 | struct dm_target_io *tio = mempool_alloc(ci->md->tio_pool, GFP_NOIO); |
f9ab94ce MP |
1101 | |
1102 | tio->io = ci->io; | |
1103 | tio->ti = ti; | |
f9ab94ce | 1104 | memset(&tio->info, 0, sizeof(tio->info)); |
9015df24 AK |
1105 | |
1106 | return tio; | |
1107 | } | |
1108 | ||
06a426ce | 1109 | static void __issue_target_request(struct clone_info *ci, struct dm_target *ti, |
a79245b3 | 1110 | unsigned request_nr, sector_t len) |
9015df24 AK |
1111 | { |
1112 | struct dm_target_io *tio = alloc_tio(ci, ti); | |
1113 | struct bio *clone; | |
1114 | ||
57cba5d3 | 1115 | tio->info.target_request_nr = request_nr; |
f9ab94ce | 1116 | |
06a426ce MS |
1117 | /* |
1118 | * Discard requests require the bio's inline iovecs be initialized. | |
1119 | * ci->bio->bi_max_vecs is BIO_INLINE_VECS anyway, for both flush | |
1120 | * and discard, so no need for concern about wasted bvec allocations. | |
1121 | */ | |
1122 | clone = bio_alloc_bioset(GFP_NOIO, ci->bio->bi_max_vecs, ci->md->bs); | |
f9ab94ce MP |
1123 | __bio_clone(clone, ci->bio); |
1124 | clone->bi_destructor = dm_bio_destructor; | |
a79245b3 MS |
1125 | if (len) { |
1126 | clone->bi_sector = ci->sector; | |
1127 | clone->bi_size = to_bytes(len); | |
1128 | } | |
f9ab94ce MP |
1129 | |
1130 | __map_bio(ti, clone, tio); | |
1131 | } | |
1132 | ||
06a426ce | 1133 | static void __issue_target_requests(struct clone_info *ci, struct dm_target *ti, |
a79245b3 | 1134 | unsigned num_requests, sector_t len) |
06a426ce MS |
1135 | { |
1136 | unsigned request_nr; | |
1137 | ||
1138 | for (request_nr = 0; request_nr < num_requests; request_nr++) | |
a79245b3 | 1139 | __issue_target_request(ci, ti, request_nr, len); |
06a426ce MS |
1140 | } |
1141 | ||
b372d360 | 1142 | static int __clone_and_map_empty_flush(struct clone_info *ci) |
f9ab94ce | 1143 | { |
06a426ce | 1144 | unsigned target_nr = 0; |
f9ab94ce MP |
1145 | struct dm_target *ti; |
1146 | ||
b372d360 | 1147 | BUG_ON(bio_has_data(ci->bio)); |
f9ab94ce | 1148 | while ((ti = dm_table_get_target(ci->map, target_nr++))) |
a79245b3 | 1149 | __issue_target_requests(ci, ti, ti->num_flush_requests, 0); |
f9ab94ce | 1150 | |
f9ab94ce MP |
1151 | return 0; |
1152 | } | |
1153 | ||
5ae89a87 MS |
1154 | /* |
1155 | * Perform all io with a single clone. | |
1156 | */ | |
1157 | static void __clone_and_map_simple(struct clone_info *ci, struct dm_target *ti) | |
1158 | { | |
1159 | struct bio *clone, *bio = ci->bio; | |
1160 | struct dm_target_io *tio; | |
1161 | ||
1162 | tio = alloc_tio(ci, ti); | |
1163 | clone = clone_bio(bio, ci->sector, ci->idx, | |
1164 | bio->bi_vcnt - ci->idx, ci->sector_count, | |
1165 | ci->md->bs); | |
1166 | __map_bio(ti, clone, tio); | |
1167 | ci->sector_count = 0; | |
1168 | } | |
1169 | ||
1170 | static int __clone_and_map_discard(struct clone_info *ci) | |
1171 | { | |
1172 | struct dm_target *ti; | |
a79245b3 | 1173 | sector_t len; |
5ae89a87 | 1174 | |
a79245b3 MS |
1175 | do { |
1176 | ti = dm_table_find_target(ci->map, ci->sector); | |
1177 | if (!dm_target_is_valid(ti)) | |
1178 | return -EIO; | |
5ae89a87 | 1179 | |
5ae89a87 | 1180 | /* |
a79245b3 MS |
1181 | * Even though the device advertised discard support, |
1182 | * reconfiguration might have changed that since the | |
1183 | * check was performed. | |
5ae89a87 | 1184 | */ |
a79245b3 MS |
1185 | if (!ti->num_discard_requests) |
1186 | return -EOPNOTSUPP; | |
5ae89a87 | 1187 | |
a79245b3 | 1188 | len = min(ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); |
06a426ce | 1189 | |
a79245b3 MS |
1190 | __issue_target_requests(ci, ti, ti->num_discard_requests, len); |
1191 | ||
1192 | ci->sector += len; | |
1193 | } while (ci->sector_count -= len); | |
5ae89a87 MS |
1194 | |
1195 | return 0; | |
1196 | } | |
1197 | ||
512875bd | 1198 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
1199 | { |
1200 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
1201 | struct dm_target *ti; |
1202 | sector_t len = 0, max; | |
028867ac | 1203 | struct dm_target_io *tio; |
1da177e4 | 1204 | |
5ae89a87 MS |
1205 | if (unlikely(bio->bi_rw & REQ_DISCARD)) |
1206 | return __clone_and_map_discard(ci); | |
1207 | ||
512875bd JN |
1208 | ti = dm_table_find_target(ci->map, ci->sector); |
1209 | if (!dm_target_is_valid(ti)) | |
1210 | return -EIO; | |
1211 | ||
56a67df7 | 1212 | max = max_io_len(ci->sector, ti); |
512875bd | 1213 | |
1da177e4 LT |
1214 | if (ci->sector_count <= max) { |
1215 | /* | |
1216 | * Optimise for the simple case where we can do all of | |
1217 | * the remaining io with a single clone. | |
1218 | */ | |
5ae89a87 | 1219 | __clone_and_map_simple(ci, ti); |
1da177e4 LT |
1220 | |
1221 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
1222 | /* | |
1223 | * There are some bvecs that don't span targets. | |
1224 | * Do as many of these as possible. | |
1225 | */ | |
1226 | int i; | |
1227 | sector_t remaining = max; | |
1228 | sector_t bv_len; | |
1229 | ||
1230 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
1231 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
1232 | ||
1233 | if (bv_len > remaining) | |
1234 | break; | |
1235 | ||
1236 | remaining -= bv_len; | |
1237 | len += bv_len; | |
1238 | } | |
1239 | ||
5ae89a87 | 1240 | tio = alloc_tio(ci, ti); |
9faf400f SB |
1241 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
1242 | ci->md->bs); | |
1da177e4 LT |
1243 | __map_bio(ti, clone, tio); |
1244 | ||
1245 | ci->sector += len; | |
1246 | ci->sector_count -= len; | |
1247 | ci->idx = i; | |
1248 | ||
1249 | } else { | |
1250 | /* | |
d2044a94 | 1251 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
1252 | */ |
1253 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
1254 | sector_t remaining = to_sector(bv->bv_len); |
1255 | unsigned int offset = 0; | |
1da177e4 | 1256 | |
d2044a94 AK |
1257 | do { |
1258 | if (offset) { | |
1259 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
1260 | if (!dm_target_is_valid(ti)) |
1261 | return -EIO; | |
1262 | ||
56a67df7 | 1263 | max = max_io_len(ci->sector, ti); |
d2044a94 AK |
1264 | } |
1265 | ||
1266 | len = min(remaining, max); | |
1267 | ||
5ae89a87 | 1268 | tio = alloc_tio(ci, ti); |
d2044a94 | 1269 | clone = split_bvec(bio, ci->sector, ci->idx, |
9faf400f SB |
1270 | bv->bv_offset + offset, len, |
1271 | ci->md->bs); | |
d2044a94 AK |
1272 | |
1273 | __map_bio(ti, clone, tio); | |
1274 | ||
1275 | ci->sector += len; | |
1276 | ci->sector_count -= len; | |
1277 | offset += to_bytes(len); | |
1278 | } while (remaining -= len); | |
1da177e4 | 1279 | |
1da177e4 LT |
1280 | ci->idx++; |
1281 | } | |
512875bd JN |
1282 | |
1283 | return 0; | |
1da177e4 LT |
1284 | } |
1285 | ||
1286 | /* | |
8a53c28d | 1287 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 1288 | */ |
f0b9a450 | 1289 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 LT |
1290 | { |
1291 | struct clone_info ci; | |
512875bd | 1292 | int error = 0; |
1da177e4 | 1293 | |
7c666411 | 1294 | ci.map = dm_get_live_table(md); |
f0b9a450 | 1295 | if (unlikely(!ci.map)) { |
6a8736d1 | 1296 | bio_io_error(bio); |
f0b9a450 MP |
1297 | return; |
1298 | } | |
692d0eb9 | 1299 | |
1da177e4 | 1300 | ci.md = md; |
1da177e4 LT |
1301 | ci.io = alloc_io(md); |
1302 | ci.io->error = 0; | |
1303 | atomic_set(&ci.io->io_count, 1); | |
1304 | ci.io->bio = bio; | |
1305 | ci.io->md = md; | |
f88fb981 | 1306 | spin_lock_init(&ci.io->endio_lock); |
1da177e4 | 1307 | ci.sector = bio->bi_sector; |
1da177e4 LT |
1308 | ci.idx = bio->bi_idx; |
1309 | ||
3eaf840e | 1310 | start_io_acct(ci.io); |
b372d360 MS |
1311 | if (bio->bi_rw & REQ_FLUSH) { |
1312 | ci.bio = &ci.md->flush_bio; | |
1313 | ci.sector_count = 0; | |
1314 | error = __clone_and_map_empty_flush(&ci); | |
1315 | /* dec_pending submits any data associated with flush */ | |
1316 | } else { | |
6a8736d1 | 1317 | ci.bio = bio; |
d87f4c14 | 1318 | ci.sector_count = bio_sectors(bio); |
b372d360 | 1319 | while (ci.sector_count && !error) |
d87f4c14 | 1320 | error = __clone_and_map(&ci); |
d87f4c14 | 1321 | } |
1da177e4 LT |
1322 | |
1323 | /* drop the extra reference count */ | |
512875bd | 1324 | dec_pending(ci.io, error); |
1da177e4 LT |
1325 | dm_table_put(ci.map); |
1326 | } | |
1327 | /*----------------------------------------------------------------- | |
1328 | * CRUD END | |
1329 | *---------------------------------------------------------------*/ | |
1330 | ||
f6fccb12 MB |
1331 | static int dm_merge_bvec(struct request_queue *q, |
1332 | struct bvec_merge_data *bvm, | |
1333 | struct bio_vec *biovec) | |
1334 | { | |
1335 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1336 | struct dm_table *map = dm_get_live_table(md); |
f6fccb12 MB |
1337 | struct dm_target *ti; |
1338 | sector_t max_sectors; | |
5037108a | 1339 | int max_size = 0; |
f6fccb12 MB |
1340 | |
1341 | if (unlikely(!map)) | |
5037108a | 1342 | goto out; |
f6fccb12 MB |
1343 | |
1344 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
1345 | if (!dm_target_is_valid(ti)) |
1346 | goto out_table; | |
f6fccb12 MB |
1347 | |
1348 | /* | |
1349 | * Find maximum amount of I/O that won't need splitting | |
1350 | */ | |
56a67df7 | 1351 | max_sectors = min(max_io_len(bvm->bi_sector, ti), |
f6fccb12 MB |
1352 | (sector_t) BIO_MAX_SECTORS); |
1353 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
1354 | if (max_size < 0) | |
1355 | max_size = 0; | |
1356 | ||
1357 | /* | |
1358 | * merge_bvec_fn() returns number of bytes | |
1359 | * it can accept at this offset | |
1360 | * max is precomputed maximal io size | |
1361 | */ | |
1362 | if (max_size && ti->type->merge) | |
1363 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
8cbeb67a MP |
1364 | /* |
1365 | * If the target doesn't support merge method and some of the devices | |
1366 | * provided their merge_bvec method (we know this by looking at | |
1367 | * queue_max_hw_sectors), then we can't allow bios with multiple vector | |
1368 | * entries. So always set max_size to 0, and the code below allows | |
1369 | * just one page. | |
1370 | */ | |
1371 | else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9) | |
1372 | ||
1373 | max_size = 0; | |
f6fccb12 | 1374 | |
b01cd5ac | 1375 | out_table: |
5037108a MP |
1376 | dm_table_put(map); |
1377 | ||
1378 | out: | |
f6fccb12 MB |
1379 | /* |
1380 | * Always allow an entire first page | |
1381 | */ | |
1382 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
1383 | max_size = biovec->bv_len; | |
1384 | ||
f6fccb12 MB |
1385 | return max_size; |
1386 | } | |
1387 | ||
1da177e4 LT |
1388 | /* |
1389 | * The request function that just remaps the bio built up by | |
1390 | * dm_merge_bvec. | |
1391 | */ | |
cec47e3d | 1392 | static int _dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 1393 | { |
12f03a49 | 1394 | int rw = bio_data_dir(bio); |
1da177e4 | 1395 | struct mapped_device *md = q->queuedata; |
c9959059 | 1396 | int cpu; |
1da177e4 | 1397 | |
2ca3310e | 1398 | down_read(&md->io_lock); |
1da177e4 | 1399 | |
074a7aca TH |
1400 | cpu = part_stat_lock(); |
1401 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
1402 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
1403 | part_stat_unlock(); | |
12f03a49 | 1404 | |
6a8736d1 TH |
1405 | /* if we're suspended, we have to queue this io for later */ |
1406 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) { | |
2ca3310e | 1407 | up_read(&md->io_lock); |
1da177e4 | 1408 | |
6a8736d1 TH |
1409 | if (bio_rw(bio) != READA) |
1410 | queue_io(md, bio); | |
1411 | else | |
54d9a1b4 | 1412 | bio_io_error(bio); |
92c63902 | 1413 | return 0; |
1da177e4 LT |
1414 | } |
1415 | ||
f0b9a450 | 1416 | __split_and_process_bio(md, bio); |
2ca3310e | 1417 | up_read(&md->io_lock); |
f0b9a450 | 1418 | return 0; |
1da177e4 LT |
1419 | } |
1420 | ||
cec47e3d KU |
1421 | static int dm_make_request(struct request_queue *q, struct bio *bio) |
1422 | { | |
1423 | struct mapped_device *md = q->queuedata; | |
1424 | ||
cec47e3d KU |
1425 | return md->saved_make_request_fn(q, bio); /* call __make_request() */ |
1426 | } | |
1427 | ||
1428 | static int dm_request_based(struct mapped_device *md) | |
1429 | { | |
1430 | return blk_queue_stackable(md->queue); | |
1431 | } | |
1432 | ||
1433 | static int dm_request(struct request_queue *q, struct bio *bio) | |
1434 | { | |
1435 | struct mapped_device *md = q->queuedata; | |
1436 | ||
1437 | if (dm_request_based(md)) | |
1438 | return dm_make_request(q, bio); | |
1439 | ||
1440 | return _dm_request(q, bio); | |
1441 | } | |
1442 | ||
1443 | void dm_dispatch_request(struct request *rq) | |
1444 | { | |
1445 | int r; | |
1446 | ||
1447 | if (blk_queue_io_stat(rq->q)) | |
1448 | rq->cmd_flags |= REQ_IO_STAT; | |
1449 | ||
1450 | rq->start_time = jiffies; | |
1451 | r = blk_insert_cloned_request(rq->q, rq); | |
1452 | if (r) | |
1453 | dm_complete_request(rq, r); | |
1454 | } | |
1455 | EXPORT_SYMBOL_GPL(dm_dispatch_request); | |
1456 | ||
1457 | static void dm_rq_bio_destructor(struct bio *bio) | |
1458 | { | |
1459 | struct dm_rq_clone_bio_info *info = bio->bi_private; | |
1460 | struct mapped_device *md = info->tio->md; | |
1461 | ||
1462 | free_bio_info(info); | |
1463 | bio_free(bio, md->bs); | |
1464 | } | |
1465 | ||
1466 | static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig, | |
1467 | void *data) | |
1468 | { | |
1469 | struct dm_rq_target_io *tio = data; | |
1470 | struct mapped_device *md = tio->md; | |
1471 | struct dm_rq_clone_bio_info *info = alloc_bio_info(md); | |
1472 | ||
1473 | if (!info) | |
1474 | return -ENOMEM; | |
1475 | ||
1476 | info->orig = bio_orig; | |
1477 | info->tio = tio; | |
1478 | bio->bi_end_io = end_clone_bio; | |
1479 | bio->bi_private = info; | |
1480 | bio->bi_destructor = dm_rq_bio_destructor; | |
1481 | ||
1482 | return 0; | |
1483 | } | |
1484 | ||
1485 | static int setup_clone(struct request *clone, struct request *rq, | |
1486 | struct dm_rq_target_io *tio) | |
1487 | { | |
d0bcb878 | 1488 | int r; |
cec47e3d | 1489 | |
29e4013d TH |
1490 | r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC, |
1491 | dm_rq_bio_constructor, tio); | |
1492 | if (r) | |
1493 | return r; | |
cec47e3d | 1494 | |
29e4013d TH |
1495 | clone->cmd = rq->cmd; |
1496 | clone->cmd_len = rq->cmd_len; | |
1497 | clone->sense = rq->sense; | |
1498 | clone->buffer = rq->buffer; | |
cec47e3d KU |
1499 | clone->end_io = end_clone_request; |
1500 | clone->end_io_data = tio; | |
1501 | ||
1502 | return 0; | |
1503 | } | |
1504 | ||
6facdaff KU |
1505 | static struct request *clone_rq(struct request *rq, struct mapped_device *md, |
1506 | gfp_t gfp_mask) | |
1507 | { | |
1508 | struct request *clone; | |
1509 | struct dm_rq_target_io *tio; | |
1510 | ||
1511 | tio = alloc_rq_tio(md, gfp_mask); | |
1512 | if (!tio) | |
1513 | return NULL; | |
1514 | ||
1515 | tio->md = md; | |
1516 | tio->ti = NULL; | |
1517 | tio->orig = rq; | |
1518 | tio->error = 0; | |
1519 | memset(&tio->info, 0, sizeof(tio->info)); | |
1520 | ||
1521 | clone = &tio->clone; | |
1522 | if (setup_clone(clone, rq, tio)) { | |
1523 | /* -ENOMEM */ | |
1524 | free_rq_tio(tio); | |
1525 | return NULL; | |
1526 | } | |
1527 | ||
1528 | return clone; | |
1529 | } | |
1530 | ||
cec47e3d KU |
1531 | /* |
1532 | * Called with the queue lock held. | |
1533 | */ | |
1534 | static int dm_prep_fn(struct request_queue *q, struct request *rq) | |
1535 | { | |
1536 | struct mapped_device *md = q->queuedata; | |
cec47e3d KU |
1537 | struct request *clone; |
1538 | ||
cec47e3d KU |
1539 | if (unlikely(rq->special)) { |
1540 | DMWARN("Already has something in rq->special."); | |
1541 | return BLKPREP_KILL; | |
1542 | } | |
1543 | ||
6facdaff KU |
1544 | clone = clone_rq(rq, md, GFP_ATOMIC); |
1545 | if (!clone) | |
cec47e3d | 1546 | return BLKPREP_DEFER; |
cec47e3d KU |
1547 | |
1548 | rq->special = clone; | |
1549 | rq->cmd_flags |= REQ_DONTPREP; | |
1550 | ||
1551 | return BLKPREP_OK; | |
1552 | } | |
1553 | ||
9eef87da KU |
1554 | /* |
1555 | * Returns: | |
1556 | * 0 : the request has been processed (not requeued) | |
1557 | * !0 : the request has been requeued | |
1558 | */ | |
1559 | static int map_request(struct dm_target *ti, struct request *clone, | |
1560 | struct mapped_device *md) | |
cec47e3d | 1561 | { |
9eef87da | 1562 | int r, requeued = 0; |
cec47e3d KU |
1563 | struct dm_rq_target_io *tio = clone->end_io_data; |
1564 | ||
1565 | /* | |
1566 | * Hold the md reference here for the in-flight I/O. | |
1567 | * We can't rely on the reference count by device opener, | |
1568 | * because the device may be closed during the request completion | |
1569 | * when all bios are completed. | |
1570 | * See the comment in rq_completed() too. | |
1571 | */ | |
1572 | dm_get(md); | |
1573 | ||
1574 | tio->ti = ti; | |
1575 | r = ti->type->map_rq(ti, clone, &tio->info); | |
1576 | switch (r) { | |
1577 | case DM_MAPIO_SUBMITTED: | |
1578 | /* The target has taken the I/O to submit by itself later */ | |
1579 | break; | |
1580 | case DM_MAPIO_REMAPPED: | |
1581 | /* The target has remapped the I/O so dispatch it */ | |
6db4ccd6 JN |
1582 | trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)), |
1583 | blk_rq_pos(tio->orig)); | |
cec47e3d KU |
1584 | dm_dispatch_request(clone); |
1585 | break; | |
1586 | case DM_MAPIO_REQUEUE: | |
1587 | /* The target wants to requeue the I/O */ | |
1588 | dm_requeue_unmapped_request(clone); | |
9eef87da | 1589 | requeued = 1; |
cec47e3d KU |
1590 | break; |
1591 | default: | |
1592 | if (r > 0) { | |
1593 | DMWARN("unimplemented target map return value: %d", r); | |
1594 | BUG(); | |
1595 | } | |
1596 | ||
1597 | /* The target wants to complete the I/O */ | |
1598 | dm_kill_unmapped_request(clone, r); | |
1599 | break; | |
1600 | } | |
9eef87da KU |
1601 | |
1602 | return requeued; | |
cec47e3d KU |
1603 | } |
1604 | ||
1605 | /* | |
1606 | * q->request_fn for request-based dm. | |
1607 | * Called with the queue lock held. | |
1608 | */ | |
1609 | static void dm_request_fn(struct request_queue *q) | |
1610 | { | |
1611 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1612 | struct dm_table *map = dm_get_live_table(md); |
cec47e3d | 1613 | struct dm_target *ti; |
b4324fee | 1614 | struct request *rq, *clone; |
29e4013d | 1615 | sector_t pos; |
cec47e3d KU |
1616 | |
1617 | /* | |
b4324fee KU |
1618 | * For suspend, check blk_queue_stopped() and increment |
1619 | * ->pending within a single queue_lock not to increment the | |
1620 | * number of in-flight I/Os after the queue is stopped in | |
1621 | * dm_suspend(). | |
cec47e3d | 1622 | */ |
7eaceacc | 1623 | while (!blk_queue_stopped(q)) { |
cec47e3d KU |
1624 | rq = blk_peek_request(q); |
1625 | if (!rq) | |
7eaceacc | 1626 | goto delay_and_out; |
cec47e3d | 1627 | |
29e4013d TH |
1628 | /* always use block 0 to find the target for flushes for now */ |
1629 | pos = 0; | |
1630 | if (!(rq->cmd_flags & REQ_FLUSH)) | |
1631 | pos = blk_rq_pos(rq); | |
1632 | ||
1633 | ti = dm_table_find_target(map, pos); | |
1634 | BUG_ON(!dm_target_is_valid(ti)); | |
d0bcb878 | 1635 | |
cec47e3d | 1636 | if (ti->type->busy && ti->type->busy(ti)) |
7eaceacc | 1637 | goto delay_and_out; |
cec47e3d KU |
1638 | |
1639 | blk_start_request(rq); | |
b4324fee KU |
1640 | clone = rq->special; |
1641 | atomic_inc(&md->pending[rq_data_dir(clone)]); | |
1642 | ||
cec47e3d | 1643 | spin_unlock(q->queue_lock); |
9eef87da KU |
1644 | if (map_request(ti, clone, md)) |
1645 | goto requeued; | |
1646 | ||
052189a2 KU |
1647 | BUG_ON(!irqs_disabled()); |
1648 | spin_lock(q->queue_lock); | |
cec47e3d KU |
1649 | } |
1650 | ||
1651 | goto out; | |
1652 | ||
9eef87da | 1653 | requeued: |
052189a2 KU |
1654 | BUG_ON(!irqs_disabled()); |
1655 | spin_lock(q->queue_lock); | |
9eef87da | 1656 | |
7eaceacc JA |
1657 | delay_and_out: |
1658 | blk_delay_queue(q, HZ / 10); | |
cec47e3d KU |
1659 | out: |
1660 | dm_table_put(map); | |
1661 | ||
1662 | return; | |
1663 | } | |
1664 | ||
1665 | int dm_underlying_device_busy(struct request_queue *q) | |
1666 | { | |
1667 | return blk_lld_busy(q); | |
1668 | } | |
1669 | EXPORT_SYMBOL_GPL(dm_underlying_device_busy); | |
1670 | ||
1671 | static int dm_lld_busy(struct request_queue *q) | |
1672 | { | |
1673 | int r; | |
1674 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1675 | struct dm_table *map = dm_get_live_table(md); |
cec47e3d KU |
1676 | |
1677 | if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) | |
1678 | r = 1; | |
1679 | else | |
1680 | r = dm_table_any_busy_target(map); | |
1681 | ||
1682 | dm_table_put(map); | |
1683 | ||
1684 | return r; | |
1685 | } | |
1686 | ||
1da177e4 LT |
1687 | static int dm_any_congested(void *congested_data, int bdi_bits) |
1688 | { | |
8a57dfc6 CS |
1689 | int r = bdi_bits; |
1690 | struct mapped_device *md = congested_data; | |
1691 | struct dm_table *map; | |
1da177e4 | 1692 | |
1eb787ec | 1693 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
7c666411 | 1694 | map = dm_get_live_table(md); |
8a57dfc6 | 1695 | if (map) { |
cec47e3d KU |
1696 | /* |
1697 | * Request-based dm cares about only own queue for | |
1698 | * the query about congestion status of request_queue | |
1699 | */ | |
1700 | if (dm_request_based(md)) | |
1701 | r = md->queue->backing_dev_info.state & | |
1702 | bdi_bits; | |
1703 | else | |
1704 | r = dm_table_any_congested(map, bdi_bits); | |
1705 | ||
8a57dfc6 CS |
1706 | dm_table_put(map); |
1707 | } | |
1708 | } | |
1709 | ||
1da177e4 LT |
1710 | return r; |
1711 | } | |
1712 | ||
1713 | /*----------------------------------------------------------------- | |
1714 | * An IDR is used to keep track of allocated minor numbers. | |
1715 | *---------------------------------------------------------------*/ | |
2b06cfff | 1716 | static void free_minor(int minor) |
1da177e4 | 1717 | { |
f32c10b0 | 1718 | spin_lock(&_minor_lock); |
1da177e4 | 1719 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1720 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1721 | } |
1722 | ||
1723 | /* | |
1724 | * See if the device with a specific minor # is free. | |
1725 | */ | |
cf13ab8e | 1726 | static int specific_minor(int minor) |
1da177e4 LT |
1727 | { |
1728 | int r, m; | |
1729 | ||
1730 | if (minor >= (1 << MINORBITS)) | |
1731 | return -EINVAL; | |
1732 | ||
62f75c2f JM |
1733 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1734 | if (!r) | |
1735 | return -ENOMEM; | |
1736 | ||
f32c10b0 | 1737 | spin_lock(&_minor_lock); |
1da177e4 LT |
1738 | |
1739 | if (idr_find(&_minor_idr, minor)) { | |
1740 | r = -EBUSY; | |
1741 | goto out; | |
1742 | } | |
1743 | ||
ba61fdd1 | 1744 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1745 | if (r) |
1da177e4 | 1746 | goto out; |
1da177e4 LT |
1747 | |
1748 | if (m != minor) { | |
1749 | idr_remove(&_minor_idr, m); | |
1750 | r = -EBUSY; | |
1751 | goto out; | |
1752 | } | |
1753 | ||
1754 | out: | |
f32c10b0 | 1755 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1756 | return r; |
1757 | } | |
1758 | ||
cf13ab8e | 1759 | static int next_free_minor(int *minor) |
1da177e4 | 1760 | { |
2b06cfff | 1761 | int r, m; |
1da177e4 | 1762 | |
1da177e4 | 1763 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1764 | if (!r) |
1765 | return -ENOMEM; | |
1766 | ||
f32c10b0 | 1767 | spin_lock(&_minor_lock); |
1da177e4 | 1768 | |
ba61fdd1 | 1769 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1770 | if (r) |
1da177e4 | 1771 | goto out; |
1da177e4 LT |
1772 | |
1773 | if (m >= (1 << MINORBITS)) { | |
1774 | idr_remove(&_minor_idr, m); | |
1775 | r = -ENOSPC; | |
1776 | goto out; | |
1777 | } | |
1778 | ||
1779 | *minor = m; | |
1780 | ||
1781 | out: | |
f32c10b0 | 1782 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1783 | return r; |
1784 | } | |
1785 | ||
83d5cde4 | 1786 | static const struct block_device_operations dm_blk_dops; |
1da177e4 | 1787 | |
53d5914f MP |
1788 | static void dm_wq_work(struct work_struct *work); |
1789 | ||
4a0b4ddf MS |
1790 | static void dm_init_md_queue(struct mapped_device *md) |
1791 | { | |
1792 | /* | |
1793 | * Request-based dm devices cannot be stacked on top of bio-based dm | |
1794 | * devices. The type of this dm device has not been decided yet. | |
1795 | * The type is decided at the first table loading time. | |
1796 | * To prevent problematic device stacking, clear the queue flag | |
1797 | * for request stacking support until then. | |
1798 | * | |
1799 | * This queue is new, so no concurrency on the queue_flags. | |
1800 | */ | |
1801 | queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue); | |
1802 | ||
1803 | md->queue->queuedata = md; | |
1804 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1805 | md->queue->backing_dev_info.congested_data = md; | |
1806 | blk_queue_make_request(md->queue, dm_request); | |
1807 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); | |
4a0b4ddf | 1808 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); |
d87f4c14 | 1809 | blk_queue_flush(md->queue, REQ_FLUSH | REQ_FUA); |
4a0b4ddf MS |
1810 | } |
1811 | ||
1da177e4 LT |
1812 | /* |
1813 | * Allocate and initialise a blank device with a given minor. | |
1814 | */ | |
2b06cfff | 1815 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1816 | { |
1817 | int r; | |
cf13ab8e | 1818 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1819 | void *old_md; |
1da177e4 LT |
1820 | |
1821 | if (!md) { | |
1822 | DMWARN("unable to allocate device, out of memory."); | |
1823 | return NULL; | |
1824 | } | |
1825 | ||
10da4f79 | 1826 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1827 | goto bad_module_get; |
10da4f79 | 1828 | |
1da177e4 | 1829 | /* get a minor number for the dev */ |
2b06cfff | 1830 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1831 | r = next_free_minor(&minor); |
2b06cfff | 1832 | else |
cf13ab8e | 1833 | r = specific_minor(minor); |
1da177e4 | 1834 | if (r < 0) |
6ed7ade8 | 1835 | goto bad_minor; |
1da177e4 | 1836 | |
a5664dad | 1837 | md->type = DM_TYPE_NONE; |
2ca3310e | 1838 | init_rwsem(&md->io_lock); |
e61290a4 | 1839 | mutex_init(&md->suspend_lock); |
a5664dad | 1840 | mutex_init(&md->type_lock); |
022c2611 | 1841 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1842 | rwlock_init(&md->map_lock); |
1843 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1844 | atomic_set(&md->open_count, 0); |
1da177e4 | 1845 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1846 | atomic_set(&md->uevent_seq, 0); |
1847 | INIT_LIST_HEAD(&md->uevent_list); | |
1848 | spin_lock_init(&md->uevent_lock); | |
1da177e4 | 1849 | |
4a0b4ddf | 1850 | md->queue = blk_alloc_queue(GFP_KERNEL); |
1da177e4 | 1851 | if (!md->queue) |
6ed7ade8 | 1852 | goto bad_queue; |
1da177e4 | 1853 | |
4a0b4ddf | 1854 | dm_init_md_queue(md); |
9faf400f | 1855 | |
1da177e4 LT |
1856 | md->disk = alloc_disk(1); |
1857 | if (!md->disk) | |
6ed7ade8 | 1858 | goto bad_disk; |
1da177e4 | 1859 | |
316d315b NK |
1860 | atomic_set(&md->pending[0], 0); |
1861 | atomic_set(&md->pending[1], 0); | |
f0b04115 | 1862 | init_waitqueue_head(&md->wait); |
53d5914f | 1863 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1864 | init_waitqueue_head(&md->eventq); |
1865 | ||
1da177e4 LT |
1866 | md->disk->major = _major; |
1867 | md->disk->first_minor = minor; | |
1868 | md->disk->fops = &dm_blk_dops; | |
1869 | md->disk->queue = md->queue; | |
1870 | md->disk->private_data = md; | |
1871 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1872 | add_disk(md->disk); | |
7e51f257 | 1873 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1874 | |
9c4376de TH |
1875 | md->wq = alloc_workqueue("kdmflush", |
1876 | WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0); | |
304f3f6a MB |
1877 | if (!md->wq) |
1878 | goto bad_thread; | |
1879 | ||
32a926da MP |
1880 | md->bdev = bdget_disk(md->disk, 0); |
1881 | if (!md->bdev) | |
1882 | goto bad_bdev; | |
1883 | ||
6a8736d1 TH |
1884 | bio_init(&md->flush_bio); |
1885 | md->flush_bio.bi_bdev = md->bdev; | |
1886 | md->flush_bio.bi_rw = WRITE_FLUSH; | |
1887 | ||
ba61fdd1 | 1888 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1889 | spin_lock(&_minor_lock); |
ba61fdd1 | 1890 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1891 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1892 | |
1893 | BUG_ON(old_md != MINOR_ALLOCED); | |
1894 | ||
1da177e4 LT |
1895 | return md; |
1896 | ||
32a926da MP |
1897 | bad_bdev: |
1898 | destroy_workqueue(md->wq); | |
304f3f6a | 1899 | bad_thread: |
03022c54 | 1900 | del_gendisk(md->disk); |
304f3f6a | 1901 | put_disk(md->disk); |
6ed7ade8 | 1902 | bad_disk: |
1312f40e | 1903 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1904 | bad_queue: |
1da177e4 | 1905 | free_minor(minor); |
6ed7ade8 | 1906 | bad_minor: |
10da4f79 | 1907 | module_put(THIS_MODULE); |
6ed7ade8 | 1908 | bad_module_get: |
1da177e4 LT |
1909 | kfree(md); |
1910 | return NULL; | |
1911 | } | |
1912 | ||
ae9da83f JN |
1913 | static void unlock_fs(struct mapped_device *md); |
1914 | ||
1da177e4 LT |
1915 | static void free_dev(struct mapped_device *md) |
1916 | { | |
f331c029 | 1917 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1918 | |
32a926da MP |
1919 | unlock_fs(md); |
1920 | bdput(md->bdev); | |
304f3f6a | 1921 | destroy_workqueue(md->wq); |
e6ee8c0b KU |
1922 | if (md->tio_pool) |
1923 | mempool_destroy(md->tio_pool); | |
1924 | if (md->io_pool) | |
1925 | mempool_destroy(md->io_pool); | |
1926 | if (md->bs) | |
1927 | bioset_free(md->bs); | |
9c47008d | 1928 | blk_integrity_unregister(md->disk); |
1da177e4 | 1929 | del_gendisk(md->disk); |
63d94e48 | 1930 | free_minor(minor); |
fba9f90e JM |
1931 | |
1932 | spin_lock(&_minor_lock); | |
1933 | md->disk->private_data = NULL; | |
1934 | spin_unlock(&_minor_lock); | |
1935 | ||
1da177e4 | 1936 | put_disk(md->disk); |
1312f40e | 1937 | blk_cleanup_queue(md->queue); |
10da4f79 | 1938 | module_put(THIS_MODULE); |
1da177e4 LT |
1939 | kfree(md); |
1940 | } | |
1941 | ||
e6ee8c0b KU |
1942 | static void __bind_mempools(struct mapped_device *md, struct dm_table *t) |
1943 | { | |
1944 | struct dm_md_mempools *p; | |
1945 | ||
1946 | if (md->io_pool && md->tio_pool && md->bs) | |
1947 | /* the md already has necessary mempools */ | |
1948 | goto out; | |
1949 | ||
1950 | p = dm_table_get_md_mempools(t); | |
1951 | BUG_ON(!p || md->io_pool || md->tio_pool || md->bs); | |
1952 | ||
1953 | md->io_pool = p->io_pool; | |
1954 | p->io_pool = NULL; | |
1955 | md->tio_pool = p->tio_pool; | |
1956 | p->tio_pool = NULL; | |
1957 | md->bs = p->bs; | |
1958 | p->bs = NULL; | |
1959 | ||
1960 | out: | |
1961 | /* mempool bind completed, now no need any mempools in the table */ | |
1962 | dm_table_free_md_mempools(t); | |
1963 | } | |
1964 | ||
1da177e4 LT |
1965 | /* |
1966 | * Bind a table to the device. | |
1967 | */ | |
1968 | static void event_callback(void *context) | |
1969 | { | |
7a8c3d3b MA |
1970 | unsigned long flags; |
1971 | LIST_HEAD(uevents); | |
1da177e4 LT |
1972 | struct mapped_device *md = (struct mapped_device *) context; |
1973 | ||
7a8c3d3b MA |
1974 | spin_lock_irqsave(&md->uevent_lock, flags); |
1975 | list_splice_init(&md->uevent_list, &uevents); | |
1976 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1977 | ||
ed9e1982 | 1978 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1979 | |
1da177e4 LT |
1980 | atomic_inc(&md->event_nr); |
1981 | wake_up(&md->eventq); | |
1982 | } | |
1983 | ||
c217649b MS |
1984 | /* |
1985 | * Protected by md->suspend_lock obtained by dm_swap_table(). | |
1986 | */ | |
4e90188b | 1987 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1988 | { |
4e90188b | 1989 | set_capacity(md->disk, size); |
1da177e4 | 1990 | |
db8fef4f | 1991 | i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1da177e4 LT |
1992 | } |
1993 | ||
042d2a9b AK |
1994 | /* |
1995 | * Returns old map, which caller must destroy. | |
1996 | */ | |
1997 | static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, | |
1998 | struct queue_limits *limits) | |
1da177e4 | 1999 | { |
042d2a9b | 2000 | struct dm_table *old_map; |
165125e1 | 2001 | struct request_queue *q = md->queue; |
1da177e4 | 2002 | sector_t size; |
523d9297 | 2003 | unsigned long flags; |
1da177e4 LT |
2004 | |
2005 | size = dm_table_get_size(t); | |
3ac51e74 DW |
2006 | |
2007 | /* | |
2008 | * Wipe any geometry if the size of the table changed. | |
2009 | */ | |
2010 | if (size != get_capacity(md->disk)) | |
2011 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
2012 | ||
32a926da | 2013 | __set_size(md, size); |
d5816876 | 2014 | |
2ca3310e AK |
2015 | dm_table_event_callback(t, event_callback, md); |
2016 | ||
e6ee8c0b KU |
2017 | /* |
2018 | * The queue hasn't been stopped yet, if the old table type wasn't | |
2019 | * for request-based during suspension. So stop it to prevent | |
2020 | * I/O mapping before resume. | |
2021 | * This must be done before setting the queue restrictions, | |
2022 | * because request-based dm may be run just after the setting. | |
2023 | */ | |
2024 | if (dm_table_request_based(t) && !blk_queue_stopped(q)) | |
2025 | stop_queue(q); | |
2026 | ||
2027 | __bind_mempools(md, t); | |
2028 | ||
523d9297 | 2029 | write_lock_irqsave(&md->map_lock, flags); |
042d2a9b | 2030 | old_map = md->map; |
1da177e4 | 2031 | md->map = t; |
754c5fc7 | 2032 | dm_table_set_restrictions(t, q, limits); |
523d9297 | 2033 | write_unlock_irqrestore(&md->map_lock, flags); |
1da177e4 | 2034 | |
042d2a9b | 2035 | return old_map; |
1da177e4 LT |
2036 | } |
2037 | ||
a7940155 AK |
2038 | /* |
2039 | * Returns unbound table for the caller to free. | |
2040 | */ | |
2041 | static struct dm_table *__unbind(struct mapped_device *md) | |
1da177e4 LT |
2042 | { |
2043 | struct dm_table *map = md->map; | |
523d9297 | 2044 | unsigned long flags; |
1da177e4 LT |
2045 | |
2046 | if (!map) | |
a7940155 | 2047 | return NULL; |
1da177e4 LT |
2048 | |
2049 | dm_table_event_callback(map, NULL, NULL); | |
523d9297 | 2050 | write_lock_irqsave(&md->map_lock, flags); |
1da177e4 | 2051 | md->map = NULL; |
523d9297 | 2052 | write_unlock_irqrestore(&md->map_lock, flags); |
a7940155 AK |
2053 | |
2054 | return map; | |
1da177e4 LT |
2055 | } |
2056 | ||
2057 | /* | |
2058 | * Constructor for a new device. | |
2059 | */ | |
2b06cfff | 2060 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
2061 | { |
2062 | struct mapped_device *md; | |
2063 | ||
2b06cfff | 2064 | md = alloc_dev(minor); |
1da177e4 LT |
2065 | if (!md) |
2066 | return -ENXIO; | |
2067 | ||
784aae73 MB |
2068 | dm_sysfs_init(md); |
2069 | ||
1da177e4 LT |
2070 | *result = md; |
2071 | return 0; | |
2072 | } | |
2073 | ||
a5664dad MS |
2074 | /* |
2075 | * Functions to manage md->type. | |
2076 | * All are required to hold md->type_lock. | |
2077 | */ | |
2078 | void dm_lock_md_type(struct mapped_device *md) | |
2079 | { | |
2080 | mutex_lock(&md->type_lock); | |
2081 | } | |
2082 | ||
2083 | void dm_unlock_md_type(struct mapped_device *md) | |
2084 | { | |
2085 | mutex_unlock(&md->type_lock); | |
2086 | } | |
2087 | ||
2088 | void dm_set_md_type(struct mapped_device *md, unsigned type) | |
2089 | { | |
2090 | md->type = type; | |
2091 | } | |
2092 | ||
2093 | unsigned dm_get_md_type(struct mapped_device *md) | |
2094 | { | |
2095 | return md->type; | |
2096 | } | |
2097 | ||
4a0b4ddf MS |
2098 | /* |
2099 | * Fully initialize a request-based queue (->elevator, ->request_fn, etc). | |
2100 | */ | |
2101 | static int dm_init_request_based_queue(struct mapped_device *md) | |
2102 | { | |
2103 | struct request_queue *q = NULL; | |
2104 | ||
2105 | if (md->queue->elevator) | |
2106 | return 1; | |
2107 | ||
2108 | /* Fully initialize the queue */ | |
2109 | q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL); | |
2110 | if (!q) | |
2111 | return 0; | |
2112 | ||
2113 | md->queue = q; | |
2114 | md->saved_make_request_fn = md->queue->make_request_fn; | |
2115 | dm_init_md_queue(md); | |
2116 | blk_queue_softirq_done(md->queue, dm_softirq_done); | |
2117 | blk_queue_prep_rq(md->queue, dm_prep_fn); | |
2118 | blk_queue_lld_busy(md->queue, dm_lld_busy); | |
4a0b4ddf MS |
2119 | |
2120 | elv_register_queue(md->queue); | |
2121 | ||
2122 | return 1; | |
2123 | } | |
2124 | ||
2125 | /* | |
2126 | * Setup the DM device's queue based on md's type | |
2127 | */ | |
2128 | int dm_setup_md_queue(struct mapped_device *md) | |
2129 | { | |
2130 | if ((dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) && | |
2131 | !dm_init_request_based_queue(md)) { | |
2132 | DMWARN("Cannot initialize queue for request-based mapped device"); | |
2133 | return -EINVAL; | |
2134 | } | |
2135 | ||
2136 | return 0; | |
2137 | } | |
2138 | ||
637842cf | 2139 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
2140 | { |
2141 | struct mapped_device *md; | |
1da177e4 LT |
2142 | unsigned minor = MINOR(dev); |
2143 | ||
2144 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
2145 | return NULL; | |
2146 | ||
f32c10b0 | 2147 | spin_lock(&_minor_lock); |
1da177e4 LT |
2148 | |
2149 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 2150 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 2151 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
abdc568b | 2152 | dm_deleting_md(md) || |
17b2f66f | 2153 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 2154 | md = NULL; |
fba9f90e JM |
2155 | goto out; |
2156 | } | |
1da177e4 | 2157 | |
fba9f90e | 2158 | out: |
f32c10b0 | 2159 | spin_unlock(&_minor_lock); |
1da177e4 | 2160 | |
637842cf DT |
2161 | return md; |
2162 | } | |
2163 | ||
d229a958 DT |
2164 | struct mapped_device *dm_get_md(dev_t dev) |
2165 | { | |
2166 | struct mapped_device *md = dm_find_md(dev); | |
2167 | ||
2168 | if (md) | |
2169 | dm_get(md); | |
2170 | ||
2171 | return md; | |
2172 | } | |
2173 | ||
9ade92a9 | 2174 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 2175 | { |
9ade92a9 | 2176 | return md->interface_ptr; |
1da177e4 LT |
2177 | } |
2178 | ||
2179 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
2180 | { | |
2181 | md->interface_ptr = ptr; | |
2182 | } | |
2183 | ||
2184 | void dm_get(struct mapped_device *md) | |
2185 | { | |
2186 | atomic_inc(&md->holders); | |
3f77316d | 2187 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1da177e4 LT |
2188 | } |
2189 | ||
72d94861 AK |
2190 | const char *dm_device_name(struct mapped_device *md) |
2191 | { | |
2192 | return md->name; | |
2193 | } | |
2194 | EXPORT_SYMBOL_GPL(dm_device_name); | |
2195 | ||
3f77316d | 2196 | static void __dm_destroy(struct mapped_device *md, bool wait) |
1da177e4 | 2197 | { |
1134e5ae | 2198 | struct dm_table *map; |
1da177e4 | 2199 | |
3f77316d | 2200 | might_sleep(); |
fba9f90e | 2201 | |
3f77316d KU |
2202 | spin_lock(&_minor_lock); |
2203 | map = dm_get_live_table(md); | |
2204 | idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); | |
2205 | set_bit(DMF_FREEING, &md->flags); | |
2206 | spin_unlock(&_minor_lock); | |
2207 | ||
2208 | if (!dm_suspended_md(md)) { | |
2209 | dm_table_presuspend_targets(map); | |
2210 | dm_table_postsuspend_targets(map); | |
1da177e4 | 2211 | } |
3f77316d KU |
2212 | |
2213 | /* | |
2214 | * Rare, but there may be I/O requests still going to complete, | |
2215 | * for example. Wait for all references to disappear. | |
2216 | * No one should increment the reference count of the mapped_device, | |
2217 | * after the mapped_device state becomes DMF_FREEING. | |
2218 | */ | |
2219 | if (wait) | |
2220 | while (atomic_read(&md->holders)) | |
2221 | msleep(1); | |
2222 | else if (atomic_read(&md->holders)) | |
2223 | DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)", | |
2224 | dm_device_name(md), atomic_read(&md->holders)); | |
2225 | ||
2226 | dm_sysfs_exit(md); | |
2227 | dm_table_put(map); | |
2228 | dm_table_destroy(__unbind(md)); | |
2229 | free_dev(md); | |
2230 | } | |
2231 | ||
2232 | void dm_destroy(struct mapped_device *md) | |
2233 | { | |
2234 | __dm_destroy(md, true); | |
2235 | } | |
2236 | ||
2237 | void dm_destroy_immediate(struct mapped_device *md) | |
2238 | { | |
2239 | __dm_destroy(md, false); | |
2240 | } | |
2241 | ||
2242 | void dm_put(struct mapped_device *md) | |
2243 | { | |
2244 | atomic_dec(&md->holders); | |
1da177e4 | 2245 | } |
79eb885c | 2246 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 2247 | |
401600df | 2248 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
2249 | { |
2250 | int r = 0; | |
b44ebeb0 MP |
2251 | DECLARE_WAITQUEUE(wait, current); |
2252 | ||
b44ebeb0 | 2253 | add_wait_queue(&md->wait, &wait); |
46125c1c MB |
2254 | |
2255 | while (1) { | |
401600df | 2256 | set_current_state(interruptible); |
46125c1c MB |
2257 | |
2258 | smp_mb(); | |
b4324fee | 2259 | if (!md_in_flight(md)) |
46125c1c MB |
2260 | break; |
2261 | ||
401600df MP |
2262 | if (interruptible == TASK_INTERRUPTIBLE && |
2263 | signal_pending(current)) { | |
46125c1c MB |
2264 | r = -EINTR; |
2265 | break; | |
2266 | } | |
2267 | ||
2268 | io_schedule(); | |
2269 | } | |
2270 | set_current_state(TASK_RUNNING); | |
2271 | ||
b44ebeb0 MP |
2272 | remove_wait_queue(&md->wait, &wait); |
2273 | ||
46125c1c MB |
2274 | return r; |
2275 | } | |
2276 | ||
1da177e4 LT |
2277 | /* |
2278 | * Process the deferred bios | |
2279 | */ | |
ef208587 | 2280 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 2281 | { |
ef208587 MP |
2282 | struct mapped_device *md = container_of(work, struct mapped_device, |
2283 | work); | |
6d6f10df | 2284 | struct bio *c; |
1da177e4 | 2285 | |
6a8736d1 | 2286 | down_read(&md->io_lock); |
ef208587 | 2287 | |
3b00b203 | 2288 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
2289 | spin_lock_irq(&md->deferred_lock); |
2290 | c = bio_list_pop(&md->deferred); | |
2291 | spin_unlock_irq(&md->deferred_lock); | |
2292 | ||
6a8736d1 | 2293 | if (!c) |
df12ee99 | 2294 | break; |
022c2611 | 2295 | |
6a8736d1 | 2296 | up_read(&md->io_lock); |
3b00b203 | 2297 | |
e6ee8c0b KU |
2298 | if (dm_request_based(md)) |
2299 | generic_make_request(c); | |
6a8736d1 TH |
2300 | else |
2301 | __split_and_process_bio(md, c); | |
3b00b203 | 2302 | |
6a8736d1 | 2303 | down_read(&md->io_lock); |
022c2611 | 2304 | } |
73d410c0 | 2305 | |
6a8736d1 | 2306 | up_read(&md->io_lock); |
1da177e4 LT |
2307 | } |
2308 | ||
9a1fb464 | 2309 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 2310 | { |
3b00b203 MP |
2311 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
2312 | smp_mb__after_clear_bit(); | |
53d5914f | 2313 | queue_work(md->wq, &md->work); |
304f3f6a MB |
2314 | } |
2315 | ||
1da177e4 | 2316 | /* |
042d2a9b | 2317 | * Swap in a new table, returning the old one for the caller to destroy. |
1da177e4 | 2318 | */ |
042d2a9b | 2319 | struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table) |
1da177e4 | 2320 | { |
042d2a9b | 2321 | struct dm_table *map = ERR_PTR(-EINVAL); |
754c5fc7 | 2322 | struct queue_limits limits; |
042d2a9b | 2323 | int r; |
1da177e4 | 2324 | |
e61290a4 | 2325 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
2326 | |
2327 | /* device must be suspended */ | |
4f186f8b | 2328 | if (!dm_suspended_md(md)) |
93c534ae | 2329 | goto out; |
1da177e4 | 2330 | |
754c5fc7 | 2331 | r = dm_calculate_queue_limits(table, &limits); |
042d2a9b AK |
2332 | if (r) { |
2333 | map = ERR_PTR(r); | |
754c5fc7 | 2334 | goto out; |
042d2a9b | 2335 | } |
754c5fc7 | 2336 | |
042d2a9b | 2337 | map = __bind(md, table, &limits); |
1da177e4 | 2338 | |
93c534ae | 2339 | out: |
e61290a4 | 2340 | mutex_unlock(&md->suspend_lock); |
042d2a9b | 2341 | return map; |
1da177e4 LT |
2342 | } |
2343 | ||
2344 | /* | |
2345 | * Functions to lock and unlock any filesystem running on the | |
2346 | * device. | |
2347 | */ | |
2ca3310e | 2348 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 2349 | { |
e39e2e95 | 2350 | int r; |
1da177e4 LT |
2351 | |
2352 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 2353 | |
db8fef4f | 2354 | md->frozen_sb = freeze_bdev(md->bdev); |
dfbe03f6 | 2355 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 2356 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
2357 | md->frozen_sb = NULL; |
2358 | return r; | |
dfbe03f6 AK |
2359 | } |
2360 | ||
aa8d7c2f AK |
2361 | set_bit(DMF_FROZEN, &md->flags); |
2362 | ||
1da177e4 LT |
2363 | return 0; |
2364 | } | |
2365 | ||
2ca3310e | 2366 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 2367 | { |
aa8d7c2f AK |
2368 | if (!test_bit(DMF_FROZEN, &md->flags)) |
2369 | return; | |
2370 | ||
db8fef4f | 2371 | thaw_bdev(md->bdev, md->frozen_sb); |
1da177e4 | 2372 | md->frozen_sb = NULL; |
aa8d7c2f | 2373 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
2374 | } |
2375 | ||
2376 | /* | |
2377 | * We need to be able to change a mapping table under a mounted | |
2378 | * filesystem. For example we might want to move some data in | |
2379 | * the background. Before the table can be swapped with | |
2380 | * dm_bind_table, dm_suspend must be called to flush any in | |
2381 | * flight bios and ensure that any further io gets deferred. | |
2382 | */ | |
cec47e3d KU |
2383 | /* |
2384 | * Suspend mechanism in request-based dm. | |
2385 | * | |
9f518b27 KU |
2386 | * 1. Flush all I/Os by lock_fs() if needed. |
2387 | * 2. Stop dispatching any I/O by stopping the request_queue. | |
2388 | * 3. Wait for all in-flight I/Os to be completed or requeued. | |
cec47e3d | 2389 | * |
9f518b27 | 2390 | * To abort suspend, start the request_queue. |
cec47e3d | 2391 | */ |
a3d77d35 | 2392 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 2393 | { |
2ca3310e | 2394 | struct dm_table *map = NULL; |
46125c1c | 2395 | int r = 0; |
a3d77d35 | 2396 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 2397 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 2398 | |
e61290a4 | 2399 | mutex_lock(&md->suspend_lock); |
2ca3310e | 2400 | |
4f186f8b | 2401 | if (dm_suspended_md(md)) { |
73d410c0 | 2402 | r = -EINVAL; |
d287483d | 2403 | goto out_unlock; |
73d410c0 | 2404 | } |
1da177e4 | 2405 | |
7c666411 | 2406 | map = dm_get_live_table(md); |
1da177e4 | 2407 | |
2e93ccc1 KU |
2408 | /* |
2409 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
2410 | * This flag is cleared before dm_suspend returns. | |
2411 | */ | |
2412 | if (noflush) | |
2413 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
2414 | ||
cf222b37 AK |
2415 | /* This does not get reverted if there's an error later. */ |
2416 | dm_table_presuspend_targets(map); | |
2417 | ||
32a926da | 2418 | /* |
9f518b27 KU |
2419 | * Flush I/O to the device. |
2420 | * Any I/O submitted after lock_fs() may not be flushed. | |
2421 | * noflush takes precedence over do_lockfs. | |
2422 | * (lock_fs() flushes I/Os and waits for them to complete.) | |
32a926da MP |
2423 | */ |
2424 | if (!noflush && do_lockfs) { | |
2425 | r = lock_fs(md); | |
2426 | if (r) | |
f431d966 | 2427 | goto out; |
aa8d7c2f | 2428 | } |
1da177e4 LT |
2429 | |
2430 | /* | |
3b00b203 MP |
2431 | * Here we must make sure that no processes are submitting requests |
2432 | * to target drivers i.e. no one may be executing | |
2433 | * __split_and_process_bio. This is called from dm_request and | |
2434 | * dm_wq_work. | |
2435 | * | |
2436 | * To get all processes out of __split_and_process_bio in dm_request, | |
2437 | * we take the write lock. To prevent any process from reentering | |
6a8736d1 TH |
2438 | * __split_and_process_bio from dm_request and quiesce the thread |
2439 | * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call | |
2440 | * flush_workqueue(md->wq). | |
1da177e4 | 2441 | */ |
2ca3310e | 2442 | down_write(&md->io_lock); |
1eb787ec | 2443 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
2ca3310e | 2444 | up_write(&md->io_lock); |
1da177e4 | 2445 | |
d0bcb878 | 2446 | /* |
29e4013d TH |
2447 | * Stop md->queue before flushing md->wq in case request-based |
2448 | * dm defers requests to md->wq from md->queue. | |
d0bcb878 | 2449 | */ |
cec47e3d | 2450 | if (dm_request_based(md)) |
9f518b27 | 2451 | stop_queue(md->queue); |
cec47e3d | 2452 | |
d0bcb878 KU |
2453 | flush_workqueue(md->wq); |
2454 | ||
1da177e4 | 2455 | /* |
3b00b203 MP |
2456 | * At this point no more requests are entering target request routines. |
2457 | * We call dm_wait_for_completion to wait for all existing requests | |
2458 | * to finish. | |
1da177e4 | 2459 | */ |
401600df | 2460 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 2461 | |
2ca3310e | 2462 | down_write(&md->io_lock); |
6d6f10df | 2463 | if (noflush) |
022c2611 | 2464 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 2465 | up_write(&md->io_lock); |
2e93ccc1 | 2466 | |
1da177e4 | 2467 | /* were we interrupted ? */ |
46125c1c | 2468 | if (r < 0) { |
9a1fb464 | 2469 | dm_queue_flush(md); |
73d410c0 | 2470 | |
cec47e3d | 2471 | if (dm_request_based(md)) |
9f518b27 | 2472 | start_queue(md->queue); |
cec47e3d | 2473 | |
2ca3310e | 2474 | unlock_fs(md); |
2e93ccc1 | 2475 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 2476 | } |
1da177e4 | 2477 | |
3b00b203 MP |
2478 | /* |
2479 | * If dm_wait_for_completion returned 0, the device is completely | |
2480 | * quiescent now. There is no request-processing activity. All new | |
2481 | * requests are being added to md->deferred list. | |
2482 | */ | |
2483 | ||
2ca3310e | 2484 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 2485 | |
4d4471cb KU |
2486 | dm_table_postsuspend_targets(map); |
2487 | ||
2ca3310e AK |
2488 | out: |
2489 | dm_table_put(map); | |
d287483d AK |
2490 | |
2491 | out_unlock: | |
e61290a4 | 2492 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 2493 | return r; |
1da177e4 LT |
2494 | } |
2495 | ||
2496 | int dm_resume(struct mapped_device *md) | |
2497 | { | |
cf222b37 | 2498 | int r = -EINVAL; |
cf222b37 | 2499 | struct dm_table *map = NULL; |
1da177e4 | 2500 | |
e61290a4 | 2501 | mutex_lock(&md->suspend_lock); |
4f186f8b | 2502 | if (!dm_suspended_md(md)) |
cf222b37 | 2503 | goto out; |
cf222b37 | 2504 | |
7c666411 | 2505 | map = dm_get_live_table(md); |
2ca3310e | 2506 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 2507 | goto out; |
1da177e4 | 2508 | |
8757b776 MB |
2509 | r = dm_table_resume_targets(map); |
2510 | if (r) | |
2511 | goto out; | |
2ca3310e | 2512 | |
9a1fb464 | 2513 | dm_queue_flush(md); |
2ca3310e | 2514 | |
cec47e3d KU |
2515 | /* |
2516 | * Flushing deferred I/Os must be done after targets are resumed | |
2517 | * so that mapping of targets can work correctly. | |
2518 | * Request-based dm is queueing the deferred I/Os in its request_queue. | |
2519 | */ | |
2520 | if (dm_request_based(md)) | |
2521 | start_queue(md->queue); | |
2522 | ||
2ca3310e AK |
2523 | unlock_fs(md); |
2524 | ||
2525 | clear_bit(DMF_SUSPENDED, &md->flags); | |
2526 | ||
cf222b37 AK |
2527 | r = 0; |
2528 | out: | |
2529 | dm_table_put(map); | |
e61290a4 | 2530 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 2531 | |
cf222b37 | 2532 | return r; |
1da177e4 LT |
2533 | } |
2534 | ||
2535 | /*----------------------------------------------------------------- | |
2536 | * Event notification. | |
2537 | *---------------------------------------------------------------*/ | |
3abf85b5 | 2538 | int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action, |
60935eb2 | 2539 | unsigned cookie) |
69267a30 | 2540 | { |
60935eb2 MB |
2541 | char udev_cookie[DM_COOKIE_LENGTH]; |
2542 | char *envp[] = { udev_cookie, NULL }; | |
2543 | ||
2544 | if (!cookie) | |
3abf85b5 | 2545 | return kobject_uevent(&disk_to_dev(md->disk)->kobj, action); |
60935eb2 MB |
2546 | else { |
2547 | snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u", | |
2548 | DM_COOKIE_ENV_VAR_NAME, cookie); | |
3abf85b5 PR |
2549 | return kobject_uevent_env(&disk_to_dev(md->disk)->kobj, |
2550 | action, envp); | |
60935eb2 | 2551 | } |
69267a30 AK |
2552 | } |
2553 | ||
7a8c3d3b MA |
2554 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
2555 | { | |
2556 | return atomic_add_return(1, &md->uevent_seq); | |
2557 | } | |
2558 | ||
1da177e4 LT |
2559 | uint32_t dm_get_event_nr(struct mapped_device *md) |
2560 | { | |
2561 | return atomic_read(&md->event_nr); | |
2562 | } | |
2563 | ||
2564 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
2565 | { | |
2566 | return wait_event_interruptible(md->eventq, | |
2567 | (event_nr != atomic_read(&md->event_nr))); | |
2568 | } | |
2569 | ||
7a8c3d3b MA |
2570 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
2571 | { | |
2572 | unsigned long flags; | |
2573 | ||
2574 | spin_lock_irqsave(&md->uevent_lock, flags); | |
2575 | list_add(elist, &md->uevent_list); | |
2576 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
2577 | } | |
2578 | ||
1da177e4 LT |
2579 | /* |
2580 | * The gendisk is only valid as long as you have a reference | |
2581 | * count on 'md'. | |
2582 | */ | |
2583 | struct gendisk *dm_disk(struct mapped_device *md) | |
2584 | { | |
2585 | return md->disk; | |
2586 | } | |
2587 | ||
784aae73 MB |
2588 | struct kobject *dm_kobject(struct mapped_device *md) |
2589 | { | |
2590 | return &md->kobj; | |
2591 | } | |
2592 | ||
2593 | /* | |
2594 | * struct mapped_device should not be exported outside of dm.c | |
2595 | * so use this check to verify that kobj is part of md structure | |
2596 | */ | |
2597 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
2598 | { | |
2599 | struct mapped_device *md; | |
2600 | ||
2601 | md = container_of(kobj, struct mapped_device, kobj); | |
2602 | if (&md->kobj != kobj) | |
2603 | return NULL; | |
2604 | ||
4d89b7b4 | 2605 | if (test_bit(DMF_FREEING, &md->flags) || |
432a212c | 2606 | dm_deleting_md(md)) |
4d89b7b4 MB |
2607 | return NULL; |
2608 | ||
784aae73 MB |
2609 | dm_get(md); |
2610 | return md; | |
2611 | } | |
2612 | ||
4f186f8b | 2613 | int dm_suspended_md(struct mapped_device *md) |
1da177e4 LT |
2614 | { |
2615 | return test_bit(DMF_SUSPENDED, &md->flags); | |
2616 | } | |
2617 | ||
64dbce58 KU |
2618 | int dm_suspended(struct dm_target *ti) |
2619 | { | |
ecdb2e25 | 2620 | return dm_suspended_md(dm_table_get_md(ti->table)); |
64dbce58 KU |
2621 | } |
2622 | EXPORT_SYMBOL_GPL(dm_suspended); | |
2623 | ||
2e93ccc1 KU |
2624 | int dm_noflush_suspending(struct dm_target *ti) |
2625 | { | |
ecdb2e25 | 2626 | return __noflush_suspending(dm_table_get_md(ti->table)); |
2e93ccc1 KU |
2627 | } |
2628 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
2629 | ||
a91a2785 | 2630 | struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity) |
e6ee8c0b KU |
2631 | { |
2632 | struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL); | |
a91a2785 | 2633 | unsigned int pool_size = (type == DM_TYPE_BIO_BASED) ? 16 : MIN_IOS; |
e6ee8c0b KU |
2634 | |
2635 | if (!pools) | |
2636 | return NULL; | |
2637 | ||
2638 | pools->io_pool = (type == DM_TYPE_BIO_BASED) ? | |
2639 | mempool_create_slab_pool(MIN_IOS, _io_cache) : | |
2640 | mempool_create_slab_pool(MIN_IOS, _rq_bio_info_cache); | |
2641 | if (!pools->io_pool) | |
2642 | goto free_pools_and_out; | |
2643 | ||
2644 | pools->tio_pool = (type == DM_TYPE_BIO_BASED) ? | |
2645 | mempool_create_slab_pool(MIN_IOS, _tio_cache) : | |
2646 | mempool_create_slab_pool(MIN_IOS, _rq_tio_cache); | |
2647 | if (!pools->tio_pool) | |
2648 | goto free_io_pool_and_out; | |
2649 | ||
a91a2785 | 2650 | pools->bs = bioset_create(pool_size, 0); |
e6ee8c0b KU |
2651 | if (!pools->bs) |
2652 | goto free_tio_pool_and_out; | |
2653 | ||
a91a2785 MP |
2654 | if (integrity && bioset_integrity_create(pools->bs, pool_size)) |
2655 | goto free_bioset_and_out; | |
2656 | ||
e6ee8c0b KU |
2657 | return pools; |
2658 | ||
a91a2785 MP |
2659 | free_bioset_and_out: |
2660 | bioset_free(pools->bs); | |
2661 | ||
e6ee8c0b KU |
2662 | free_tio_pool_and_out: |
2663 | mempool_destroy(pools->tio_pool); | |
2664 | ||
2665 | free_io_pool_and_out: | |
2666 | mempool_destroy(pools->io_pool); | |
2667 | ||
2668 | free_pools_and_out: | |
2669 | kfree(pools); | |
2670 | ||
2671 | return NULL; | |
2672 | } | |
2673 | ||
2674 | void dm_free_md_mempools(struct dm_md_mempools *pools) | |
2675 | { | |
2676 | if (!pools) | |
2677 | return; | |
2678 | ||
2679 | if (pools->io_pool) | |
2680 | mempool_destroy(pools->io_pool); | |
2681 | ||
2682 | if (pools->tio_pool) | |
2683 | mempool_destroy(pools->tio_pool); | |
2684 | ||
2685 | if (pools->bs) | |
2686 | bioset_free(pools->bs); | |
2687 | ||
2688 | kfree(pools); | |
2689 | } | |
2690 | ||
83d5cde4 | 2691 | static const struct block_device_operations dm_blk_dops = { |
1da177e4 LT |
2692 | .open = dm_blk_open, |
2693 | .release = dm_blk_close, | |
aa129a22 | 2694 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 2695 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
2696 | .owner = THIS_MODULE |
2697 | }; | |
2698 | ||
2699 | EXPORT_SYMBOL(dm_get_mapinfo); | |
2700 | ||
2701 | /* | |
2702 | * module hooks | |
2703 | */ | |
2704 | module_init(dm_init); | |
2705 | module_exit(dm_exit); | |
2706 | ||
2707 | module_param(major, uint, 0); | |
2708 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
2709 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
2710 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
2711 | MODULE_LICENSE("GPL"); |