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> | |
6e9624b8 | 18 | #include <linux/smp_lock.h> |
1da177e4 LT |
19 | #include <linux/mempool.h> |
20 | #include <linux/slab.h> | |
21 | #include <linux/idr.h> | |
3ac51e74 | 22 | #include <linux/hdreg.h> |
3f77316d | 23 | #include <linux/delay.h> |
55782138 LZ |
24 | |
25 | #include <trace/events/block.h> | |
1da177e4 | 26 | |
72d94861 AK |
27 | #define DM_MSG_PREFIX "core" |
28 | ||
60935eb2 MB |
29 | /* |
30 | * Cookies are numeric values sent with CHANGE and REMOVE | |
31 | * uevents while resuming, removing or renaming the device. | |
32 | */ | |
33 | #define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE" | |
34 | #define DM_COOKIE_LENGTH 24 | |
35 | ||
1da177e4 LT |
36 | static const char *_name = DM_NAME; |
37 | ||
38 | static unsigned int major = 0; | |
39 | static unsigned int _major = 0; | |
40 | ||
f32c10b0 | 41 | static DEFINE_SPINLOCK(_minor_lock); |
1da177e4 | 42 | /* |
8fbf26ad | 43 | * For bio-based dm. |
1da177e4 LT |
44 | * One of these is allocated per bio. |
45 | */ | |
46 | struct dm_io { | |
47 | struct mapped_device *md; | |
48 | int error; | |
1da177e4 | 49 | atomic_t io_count; |
6ae2fa67 | 50 | struct bio *bio; |
3eaf840e | 51 | unsigned long start_time; |
f88fb981 | 52 | spinlock_t endio_lock; |
1da177e4 LT |
53 | }; |
54 | ||
55 | /* | |
8fbf26ad | 56 | * For bio-based dm. |
1da177e4 LT |
57 | * One of these is allocated per target within a bio. Hopefully |
58 | * this will be simplified out one day. | |
59 | */ | |
028867ac | 60 | struct dm_target_io { |
1da177e4 LT |
61 | struct dm_io *io; |
62 | struct dm_target *ti; | |
63 | union map_info info; | |
64 | }; | |
65 | ||
8fbf26ad KU |
66 | /* |
67 | * For request-based dm. | |
68 | * One of these is allocated per request. | |
69 | */ | |
70 | struct dm_rq_target_io { | |
71 | struct mapped_device *md; | |
72 | struct dm_target *ti; | |
73 | struct request *orig, clone; | |
74 | int error; | |
75 | union map_info info; | |
76 | }; | |
77 | ||
78 | /* | |
79 | * For request-based dm. | |
80 | * One of these is allocated per bio. | |
81 | */ | |
82 | struct dm_rq_clone_bio_info { | |
83 | struct bio *orig; | |
cec47e3d | 84 | struct dm_rq_target_io *tio; |
8fbf26ad KU |
85 | }; |
86 | ||
1da177e4 LT |
87 | union map_info *dm_get_mapinfo(struct bio *bio) |
88 | { | |
17b2f66f | 89 | if (bio && bio->bi_private) |
028867ac | 90 | return &((struct dm_target_io *)bio->bi_private)->info; |
17b2f66f | 91 | return NULL; |
1da177e4 LT |
92 | } |
93 | ||
cec47e3d KU |
94 | union map_info *dm_get_rq_mapinfo(struct request *rq) |
95 | { | |
96 | if (rq && rq->end_io_data) | |
97 | return &((struct dm_rq_target_io *)rq->end_io_data)->info; | |
98 | return NULL; | |
99 | } | |
100 | EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo); | |
101 | ||
ba61fdd1 JM |
102 | #define MINOR_ALLOCED ((void *)-1) |
103 | ||
1da177e4 LT |
104 | /* |
105 | * Bits for the md->flags field. | |
106 | */ | |
1eb787ec | 107 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
1da177e4 | 108 | #define DMF_SUSPENDED 1 |
aa8d7c2f | 109 | #define DMF_FROZEN 2 |
fba9f90e | 110 | #define DMF_FREEING 3 |
5c6bd75d | 111 | #define DMF_DELETING 4 |
2e93ccc1 | 112 | #define DMF_NOFLUSH_SUSPENDING 5 |
1da177e4 | 113 | |
304f3f6a MB |
114 | /* |
115 | * Work processed by per-device workqueue. | |
116 | */ | |
1da177e4 | 117 | struct mapped_device { |
2ca3310e | 118 | struct rw_semaphore io_lock; |
e61290a4 | 119 | struct mutex suspend_lock; |
1da177e4 LT |
120 | rwlock_t map_lock; |
121 | atomic_t holders; | |
5c6bd75d | 122 | atomic_t open_count; |
1da177e4 LT |
123 | |
124 | unsigned long flags; | |
125 | ||
165125e1 | 126 | struct request_queue *queue; |
a5664dad | 127 | unsigned type; |
4a0b4ddf | 128 | /* Protect queue and type against concurrent access. */ |
a5664dad MS |
129 | struct mutex type_lock; |
130 | ||
1da177e4 | 131 | struct gendisk *disk; |
7e51f257 | 132 | char name[16]; |
1da177e4 LT |
133 | |
134 | void *interface_ptr; | |
135 | ||
136 | /* | |
137 | * A list of ios that arrived while we were suspended. | |
138 | */ | |
316d315b | 139 | atomic_t pending[2]; |
1da177e4 | 140 | wait_queue_head_t wait; |
53d5914f | 141 | struct work_struct work; |
74859364 | 142 | struct bio_list deferred; |
022c2611 | 143 | spinlock_t deferred_lock; |
1da177e4 | 144 | |
d0bcb878 | 145 | /* |
29e4013d | 146 | * Processing queue (flush) |
304f3f6a MB |
147 | */ |
148 | struct workqueue_struct *wq; | |
149 | ||
1da177e4 LT |
150 | /* |
151 | * The current mapping. | |
152 | */ | |
153 | struct dm_table *map; | |
154 | ||
155 | /* | |
156 | * io objects are allocated from here. | |
157 | */ | |
158 | mempool_t *io_pool; | |
159 | mempool_t *tio_pool; | |
160 | ||
9faf400f SB |
161 | struct bio_set *bs; |
162 | ||
1da177e4 LT |
163 | /* |
164 | * Event handling. | |
165 | */ | |
166 | atomic_t event_nr; | |
167 | wait_queue_head_t eventq; | |
7a8c3d3b MA |
168 | atomic_t uevent_seq; |
169 | struct list_head uevent_list; | |
170 | spinlock_t uevent_lock; /* Protect access to uevent_list */ | |
1da177e4 LT |
171 | |
172 | /* | |
173 | * freeze/thaw support require holding onto a super block | |
174 | */ | |
175 | struct super_block *frozen_sb; | |
db8fef4f | 176 | struct block_device *bdev; |
3ac51e74 DW |
177 | |
178 | /* forced geometry settings */ | |
179 | struct hd_geometry geometry; | |
784aae73 | 180 | |
cec47e3d KU |
181 | /* For saving the address of __make_request for request based dm */ |
182 | make_request_fn *saved_make_request_fn; | |
183 | ||
784aae73 MB |
184 | /* sysfs handle */ |
185 | struct kobject kobj; | |
52b1fd5a | 186 | |
d87f4c14 TH |
187 | /* zero-length flush that will be cloned and submitted to targets */ |
188 | struct bio flush_bio; | |
1da177e4 LT |
189 | }; |
190 | ||
e6ee8c0b KU |
191 | /* |
192 | * For mempools pre-allocation at the table loading time. | |
193 | */ | |
194 | struct dm_md_mempools { | |
195 | mempool_t *io_pool; | |
196 | mempool_t *tio_pool; | |
197 | struct bio_set *bs; | |
198 | }; | |
199 | ||
1da177e4 | 200 | #define MIN_IOS 256 |
e18b890b CL |
201 | static struct kmem_cache *_io_cache; |
202 | static struct kmem_cache *_tio_cache; | |
8fbf26ad KU |
203 | static struct kmem_cache *_rq_tio_cache; |
204 | static struct kmem_cache *_rq_bio_info_cache; | |
1da177e4 | 205 | |
1da177e4 LT |
206 | static int __init local_init(void) |
207 | { | |
51157b4a | 208 | int r = -ENOMEM; |
1da177e4 | 209 | |
1da177e4 | 210 | /* allocate a slab for the dm_ios */ |
028867ac | 211 | _io_cache = KMEM_CACHE(dm_io, 0); |
1da177e4 | 212 | if (!_io_cache) |
51157b4a | 213 | return r; |
1da177e4 LT |
214 | |
215 | /* allocate a slab for the target ios */ | |
028867ac | 216 | _tio_cache = KMEM_CACHE(dm_target_io, 0); |
51157b4a KU |
217 | if (!_tio_cache) |
218 | goto out_free_io_cache; | |
1da177e4 | 219 | |
8fbf26ad KU |
220 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
221 | if (!_rq_tio_cache) | |
222 | goto out_free_tio_cache; | |
223 | ||
224 | _rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0); | |
225 | if (!_rq_bio_info_cache) | |
226 | goto out_free_rq_tio_cache; | |
227 | ||
51e5b2bd | 228 | r = dm_uevent_init(); |
51157b4a | 229 | if (r) |
8fbf26ad | 230 | goto out_free_rq_bio_info_cache; |
51e5b2bd | 231 | |
1da177e4 LT |
232 | _major = major; |
233 | r = register_blkdev(_major, _name); | |
51157b4a KU |
234 | if (r < 0) |
235 | goto out_uevent_exit; | |
1da177e4 LT |
236 | |
237 | if (!_major) | |
238 | _major = r; | |
239 | ||
240 | return 0; | |
51157b4a KU |
241 | |
242 | out_uevent_exit: | |
243 | dm_uevent_exit(); | |
8fbf26ad KU |
244 | out_free_rq_bio_info_cache: |
245 | kmem_cache_destroy(_rq_bio_info_cache); | |
246 | out_free_rq_tio_cache: | |
247 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
248 | out_free_tio_cache: |
249 | kmem_cache_destroy(_tio_cache); | |
250 | out_free_io_cache: | |
251 | kmem_cache_destroy(_io_cache); | |
252 | ||
253 | return r; | |
1da177e4 LT |
254 | } |
255 | ||
256 | static void local_exit(void) | |
257 | { | |
8fbf26ad KU |
258 | kmem_cache_destroy(_rq_bio_info_cache); |
259 | kmem_cache_destroy(_rq_tio_cache); | |
1da177e4 LT |
260 | kmem_cache_destroy(_tio_cache); |
261 | kmem_cache_destroy(_io_cache); | |
00d59405 | 262 | unregister_blkdev(_major, _name); |
51e5b2bd | 263 | dm_uevent_exit(); |
1da177e4 LT |
264 | |
265 | _major = 0; | |
266 | ||
267 | DMINFO("cleaned up"); | |
268 | } | |
269 | ||
b9249e55 | 270 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
271 | local_init, |
272 | dm_target_init, | |
273 | dm_linear_init, | |
274 | dm_stripe_init, | |
952b3557 | 275 | dm_io_init, |
945fa4d2 | 276 | dm_kcopyd_init, |
1da177e4 LT |
277 | dm_interface_init, |
278 | }; | |
279 | ||
b9249e55 | 280 | static void (*_exits[])(void) = { |
1da177e4 LT |
281 | local_exit, |
282 | dm_target_exit, | |
283 | dm_linear_exit, | |
284 | dm_stripe_exit, | |
952b3557 | 285 | dm_io_exit, |
945fa4d2 | 286 | dm_kcopyd_exit, |
1da177e4 LT |
287 | dm_interface_exit, |
288 | }; | |
289 | ||
290 | static int __init dm_init(void) | |
291 | { | |
292 | const int count = ARRAY_SIZE(_inits); | |
293 | ||
294 | int r, i; | |
295 | ||
296 | for (i = 0; i < count; i++) { | |
297 | r = _inits[i](); | |
298 | if (r) | |
299 | goto bad; | |
300 | } | |
301 | ||
302 | return 0; | |
303 | ||
304 | bad: | |
305 | while (i--) | |
306 | _exits[i](); | |
307 | ||
308 | return r; | |
309 | } | |
310 | ||
311 | static void __exit dm_exit(void) | |
312 | { | |
313 | int i = ARRAY_SIZE(_exits); | |
314 | ||
315 | while (i--) | |
316 | _exits[i](); | |
317 | } | |
318 | ||
319 | /* | |
320 | * Block device functions | |
321 | */ | |
432a212c MA |
322 | int dm_deleting_md(struct mapped_device *md) |
323 | { | |
324 | return test_bit(DMF_DELETING, &md->flags); | |
325 | } | |
326 | ||
fe5f9f2c | 327 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
328 | { |
329 | struct mapped_device *md; | |
330 | ||
6e9624b8 | 331 | lock_kernel(); |
fba9f90e JM |
332 | spin_lock(&_minor_lock); |
333 | ||
fe5f9f2c | 334 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
335 | if (!md) |
336 | goto out; | |
337 | ||
5c6bd75d | 338 | if (test_bit(DMF_FREEING, &md->flags) || |
432a212c | 339 | dm_deleting_md(md)) { |
fba9f90e JM |
340 | md = NULL; |
341 | goto out; | |
342 | } | |
343 | ||
1da177e4 | 344 | dm_get(md); |
5c6bd75d | 345 | atomic_inc(&md->open_count); |
fba9f90e JM |
346 | |
347 | out: | |
348 | spin_unlock(&_minor_lock); | |
6e9624b8 | 349 | unlock_kernel(); |
fba9f90e JM |
350 | |
351 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
352 | } |
353 | ||
fe5f9f2c | 354 | static int dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 355 | { |
fe5f9f2c | 356 | struct mapped_device *md = disk->private_data; |
6e9624b8 AB |
357 | |
358 | lock_kernel(); | |
5c6bd75d | 359 | atomic_dec(&md->open_count); |
1da177e4 | 360 | dm_put(md); |
6e9624b8 AB |
361 | unlock_kernel(); |
362 | ||
1da177e4 LT |
363 | return 0; |
364 | } | |
365 | ||
5c6bd75d AK |
366 | int dm_open_count(struct mapped_device *md) |
367 | { | |
368 | return atomic_read(&md->open_count); | |
369 | } | |
370 | ||
371 | /* | |
372 | * Guarantees nothing is using the device before it's deleted. | |
373 | */ | |
374 | int dm_lock_for_deletion(struct mapped_device *md) | |
375 | { | |
376 | int r = 0; | |
377 | ||
378 | spin_lock(&_minor_lock); | |
379 | ||
380 | if (dm_open_count(md)) | |
381 | r = -EBUSY; | |
382 | else | |
383 | set_bit(DMF_DELETING, &md->flags); | |
384 | ||
385 | spin_unlock(&_minor_lock); | |
386 | ||
387 | return r; | |
388 | } | |
389 | ||
3ac51e74 DW |
390 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
391 | { | |
392 | struct mapped_device *md = bdev->bd_disk->private_data; | |
393 | ||
394 | return dm_get_geometry(md, geo); | |
395 | } | |
396 | ||
fe5f9f2c | 397 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, |
aa129a22 MB |
398 | unsigned int cmd, unsigned long arg) |
399 | { | |
fe5f9f2c | 400 | struct mapped_device *md = bdev->bd_disk->private_data; |
7c666411 | 401 | struct dm_table *map = dm_get_live_table(md); |
aa129a22 MB |
402 | struct dm_target *tgt; |
403 | int r = -ENOTTY; | |
404 | ||
aa129a22 MB |
405 | if (!map || !dm_table_get_size(map)) |
406 | goto out; | |
407 | ||
408 | /* We only support devices that have a single target */ | |
409 | if (dm_table_get_num_targets(map) != 1) | |
410 | goto out; | |
411 | ||
412 | tgt = dm_table_get_target(map, 0); | |
413 | ||
4f186f8b | 414 | if (dm_suspended_md(md)) { |
aa129a22 MB |
415 | r = -EAGAIN; |
416 | goto out; | |
417 | } | |
418 | ||
419 | if (tgt->type->ioctl) | |
647b3d00 | 420 | r = tgt->type->ioctl(tgt, cmd, arg); |
aa129a22 MB |
421 | |
422 | out: | |
423 | dm_table_put(map); | |
424 | ||
aa129a22 MB |
425 | return r; |
426 | } | |
427 | ||
028867ac | 428 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 LT |
429 | { |
430 | return mempool_alloc(md->io_pool, GFP_NOIO); | |
431 | } | |
432 | ||
028867ac | 433 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 LT |
434 | { |
435 | mempool_free(io, md->io_pool); | |
436 | } | |
437 | ||
028867ac | 438 | static void free_tio(struct mapped_device *md, struct dm_target_io *tio) |
1da177e4 LT |
439 | { |
440 | mempool_free(tio, md->tio_pool); | |
441 | } | |
442 | ||
08885643 KU |
443 | static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md, |
444 | gfp_t gfp_mask) | |
cec47e3d | 445 | { |
08885643 | 446 | return mempool_alloc(md->tio_pool, gfp_mask); |
cec47e3d KU |
447 | } |
448 | ||
449 | static void free_rq_tio(struct dm_rq_target_io *tio) | |
450 | { | |
451 | mempool_free(tio, tio->md->tio_pool); | |
452 | } | |
453 | ||
454 | static struct dm_rq_clone_bio_info *alloc_bio_info(struct mapped_device *md) | |
455 | { | |
456 | return mempool_alloc(md->io_pool, GFP_ATOMIC); | |
457 | } | |
458 | ||
459 | static void free_bio_info(struct dm_rq_clone_bio_info *info) | |
460 | { | |
461 | mempool_free(info, info->tio->md->io_pool); | |
462 | } | |
463 | ||
90abb8c4 KU |
464 | static int md_in_flight(struct mapped_device *md) |
465 | { | |
466 | return atomic_read(&md->pending[READ]) + | |
467 | atomic_read(&md->pending[WRITE]); | |
468 | } | |
469 | ||
3eaf840e JNN |
470 | static void start_io_acct(struct dm_io *io) |
471 | { | |
472 | struct mapped_device *md = io->md; | |
c9959059 | 473 | int cpu; |
316d315b | 474 | int rw = bio_data_dir(io->bio); |
3eaf840e JNN |
475 | |
476 | io->start_time = jiffies; | |
477 | ||
074a7aca TH |
478 | cpu = part_stat_lock(); |
479 | part_round_stats(cpu, &dm_disk(md)->part0); | |
480 | part_stat_unlock(); | |
316d315b | 481 | dm_disk(md)->part0.in_flight[rw] = atomic_inc_return(&md->pending[rw]); |
3eaf840e JNN |
482 | } |
483 | ||
d221d2e7 | 484 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
485 | { |
486 | struct mapped_device *md = io->md; | |
487 | struct bio *bio = io->bio; | |
488 | unsigned long duration = jiffies - io->start_time; | |
c9959059 | 489 | int pending, cpu; |
3eaf840e JNN |
490 | int rw = bio_data_dir(bio); |
491 | ||
074a7aca TH |
492 | cpu = part_stat_lock(); |
493 | part_round_stats(cpu, &dm_disk(md)->part0); | |
494 | part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration); | |
495 | part_stat_unlock(); | |
3eaf840e | 496 | |
af7e466a MP |
497 | /* |
498 | * After this is decremented the bio must not be touched if it is | |
d87f4c14 | 499 | * a flush. |
af7e466a | 500 | */ |
316d315b NK |
501 | dm_disk(md)->part0.in_flight[rw] = pending = |
502 | atomic_dec_return(&md->pending[rw]); | |
503 | pending += atomic_read(&md->pending[rw^0x1]); | |
3eaf840e | 504 | |
d221d2e7 MP |
505 | /* nudge anyone waiting on suspend queue */ |
506 | if (!pending) | |
507 | wake_up(&md->wait); | |
3eaf840e JNN |
508 | } |
509 | ||
1da177e4 LT |
510 | /* |
511 | * Add the bio to the list of deferred io. | |
512 | */ | |
92c63902 | 513 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 514 | { |
05447420 KU |
515 | unsigned long flags; |
516 | ||
517 | spin_lock_irqsave(&md->deferred_lock, flags); | |
1da177e4 | 518 | bio_list_add(&md->deferred, bio); |
05447420 | 519 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
6a8736d1 | 520 | queue_work(md->wq, &md->work); |
1da177e4 LT |
521 | } |
522 | ||
523 | /* | |
524 | * Everyone (including functions in this file), should use this | |
525 | * function to access the md->map field, and make sure they call | |
526 | * dm_table_put() when finished. | |
527 | */ | |
7c666411 | 528 | struct dm_table *dm_get_live_table(struct mapped_device *md) |
1da177e4 LT |
529 | { |
530 | struct dm_table *t; | |
523d9297 | 531 | unsigned long flags; |
1da177e4 | 532 | |
523d9297 | 533 | read_lock_irqsave(&md->map_lock, flags); |
1da177e4 LT |
534 | t = md->map; |
535 | if (t) | |
536 | dm_table_get(t); | |
523d9297 | 537 | read_unlock_irqrestore(&md->map_lock, flags); |
1da177e4 LT |
538 | |
539 | return t; | |
540 | } | |
541 | ||
3ac51e74 DW |
542 | /* |
543 | * Get the geometry associated with a dm device | |
544 | */ | |
545 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
546 | { | |
547 | *geo = md->geometry; | |
548 | ||
549 | return 0; | |
550 | } | |
551 | ||
552 | /* | |
553 | * Set the geometry of a device. | |
554 | */ | |
555 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
556 | { | |
557 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
558 | ||
559 | if (geo->start > sz) { | |
560 | DMWARN("Start sector is beyond the geometry limits."); | |
561 | return -EINVAL; | |
562 | } | |
563 | ||
564 | md->geometry = *geo; | |
565 | ||
566 | return 0; | |
567 | } | |
568 | ||
1da177e4 LT |
569 | /*----------------------------------------------------------------- |
570 | * CRUD START: | |
571 | * A more elegant soln is in the works that uses the queue | |
572 | * merge fn, unfortunately there are a couple of changes to | |
573 | * the block layer that I want to make for this. So in the | |
574 | * interests of getting something for people to use I give | |
575 | * you this clearly demarcated crap. | |
576 | *---------------------------------------------------------------*/ | |
577 | ||
2e93ccc1 KU |
578 | static int __noflush_suspending(struct mapped_device *md) |
579 | { | |
580 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
581 | } | |
582 | ||
1da177e4 LT |
583 | /* |
584 | * Decrements the number of outstanding ios that a bio has been | |
585 | * cloned into, completing the original io if necc. | |
586 | */ | |
858119e1 | 587 | static void dec_pending(struct dm_io *io, int error) |
1da177e4 | 588 | { |
2e93ccc1 | 589 | unsigned long flags; |
b35f8caa MB |
590 | int io_error; |
591 | struct bio *bio; | |
592 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
593 | |
594 | /* Push-back supersedes any I/O errors */ | |
f88fb981 KU |
595 | if (unlikely(error)) { |
596 | spin_lock_irqsave(&io->endio_lock, flags); | |
597 | if (!(io->error > 0 && __noflush_suspending(md))) | |
598 | io->error = error; | |
599 | spin_unlock_irqrestore(&io->endio_lock, flags); | |
600 | } | |
1da177e4 LT |
601 | |
602 | if (atomic_dec_and_test(&io->io_count)) { | |
2e93ccc1 KU |
603 | if (io->error == DM_ENDIO_REQUEUE) { |
604 | /* | |
605 | * Target requested pushing back the I/O. | |
2e93ccc1 | 606 | */ |
022c2611 | 607 | spin_lock_irqsave(&md->deferred_lock, flags); |
6a8736d1 TH |
608 | if (__noflush_suspending(md)) |
609 | bio_list_add_head(&md->deferred, io->bio); | |
610 | else | |
2e93ccc1 KU |
611 | /* noflush suspend was interrupted. */ |
612 | io->error = -EIO; | |
022c2611 | 613 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
614 | } |
615 | ||
b35f8caa MB |
616 | io_error = io->error; |
617 | bio = io->bio; | |
6a8736d1 TH |
618 | end_io_acct(io); |
619 | free_io(md, io); | |
620 | ||
621 | if (io_error == DM_ENDIO_REQUEUE) | |
622 | return; | |
2e93ccc1 | 623 | |
6a8736d1 TH |
624 | if (!(bio->bi_rw & REQ_FLUSH) || !bio->bi_size) { |
625 | trace_block_bio_complete(md->queue, bio); | |
626 | bio_endio(bio, io_error); | |
627 | } else { | |
af7e466a | 628 | /* |
6a8736d1 TH |
629 | * Preflush done for flush with data, reissue |
630 | * without REQ_FLUSH. | |
af7e466a | 631 | */ |
6a8736d1 TH |
632 | bio->bi_rw &= ~REQ_FLUSH; |
633 | queue_io(md, bio); | |
b35f8caa | 634 | } |
1da177e4 LT |
635 | } |
636 | } | |
637 | ||
6712ecf8 | 638 | static void clone_endio(struct bio *bio, int error) |
1da177e4 LT |
639 | { |
640 | int r = 0; | |
028867ac | 641 | struct dm_target_io *tio = bio->bi_private; |
b35f8caa | 642 | struct dm_io *io = tio->io; |
9faf400f | 643 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
644 | dm_endio_fn endio = tio->ti->type->end_io; |
645 | ||
1da177e4 LT |
646 | if (!bio_flagged(bio, BIO_UPTODATE) && !error) |
647 | error = -EIO; | |
648 | ||
649 | if (endio) { | |
650 | r = endio(tio->ti, bio, error, &tio->info); | |
2e93ccc1 KU |
651 | if (r < 0 || r == DM_ENDIO_REQUEUE) |
652 | /* | |
653 | * error and requeue request are handled | |
654 | * in dec_pending(). | |
655 | */ | |
1da177e4 | 656 | error = r; |
45cbcd79 KU |
657 | else if (r == DM_ENDIO_INCOMPLETE) |
658 | /* The target will handle the io */ | |
6712ecf8 | 659 | return; |
45cbcd79 KU |
660 | else if (r) { |
661 | DMWARN("unimplemented target endio return value: %d", r); | |
662 | BUG(); | |
663 | } | |
1da177e4 LT |
664 | } |
665 | ||
9faf400f SB |
666 | /* |
667 | * Store md for cleanup instead of tio which is about to get freed. | |
668 | */ | |
669 | bio->bi_private = md->bs; | |
670 | ||
9faf400f | 671 | free_tio(md, tio); |
b35f8caa MB |
672 | bio_put(bio); |
673 | dec_pending(io, error); | |
1da177e4 LT |
674 | } |
675 | ||
cec47e3d KU |
676 | /* |
677 | * Partial completion handling for request-based dm | |
678 | */ | |
679 | static void end_clone_bio(struct bio *clone, int error) | |
680 | { | |
681 | struct dm_rq_clone_bio_info *info = clone->bi_private; | |
682 | struct dm_rq_target_io *tio = info->tio; | |
683 | struct bio *bio = info->orig; | |
684 | unsigned int nr_bytes = info->orig->bi_size; | |
685 | ||
686 | bio_put(clone); | |
687 | ||
688 | if (tio->error) | |
689 | /* | |
690 | * An error has already been detected on the request. | |
691 | * Once error occurred, just let clone->end_io() handle | |
692 | * the remainder. | |
693 | */ | |
694 | return; | |
695 | else if (error) { | |
696 | /* | |
697 | * Don't notice the error to the upper layer yet. | |
698 | * The error handling decision is made by the target driver, | |
699 | * when the request is completed. | |
700 | */ | |
701 | tio->error = error; | |
702 | return; | |
703 | } | |
704 | ||
705 | /* | |
706 | * I/O for the bio successfully completed. | |
707 | * Notice the data completion to the upper layer. | |
708 | */ | |
709 | ||
710 | /* | |
711 | * bios are processed from the head of the list. | |
712 | * So the completing bio should always be rq->bio. | |
713 | * If it's not, something wrong is happening. | |
714 | */ | |
715 | if (tio->orig->bio != bio) | |
716 | DMERR("bio completion is going in the middle of the request"); | |
717 | ||
718 | /* | |
719 | * Update the original request. | |
720 | * Do not use blk_end_request() here, because it may complete | |
721 | * the original request before the clone, and break the ordering. | |
722 | */ | |
723 | blk_update_request(tio->orig, 0, nr_bytes); | |
724 | } | |
725 | ||
726 | /* | |
727 | * Don't touch any member of the md after calling this function because | |
728 | * the md may be freed in dm_put() at the end of this function. | |
729 | * Or do dm_get() before calling this function and dm_put() later. | |
730 | */ | |
b4324fee | 731 | static void rq_completed(struct mapped_device *md, int rw, int run_queue) |
cec47e3d | 732 | { |
b4324fee | 733 | atomic_dec(&md->pending[rw]); |
cec47e3d KU |
734 | |
735 | /* nudge anyone waiting on suspend queue */ | |
b4324fee | 736 | if (!md_in_flight(md)) |
cec47e3d KU |
737 | wake_up(&md->wait); |
738 | ||
739 | if (run_queue) | |
b4324fee | 740 | blk_run_queue(md->queue); |
cec47e3d KU |
741 | |
742 | /* | |
743 | * dm_put() must be at the end of this function. See the comment above | |
744 | */ | |
745 | dm_put(md); | |
746 | } | |
747 | ||
a77e28c7 KU |
748 | static void free_rq_clone(struct request *clone) |
749 | { | |
750 | struct dm_rq_target_io *tio = clone->end_io_data; | |
751 | ||
752 | blk_rq_unprep_clone(clone); | |
753 | free_rq_tio(tio); | |
754 | } | |
755 | ||
980691e5 KU |
756 | /* |
757 | * Complete the clone and the original request. | |
758 | * Must be called without queue lock. | |
759 | */ | |
760 | static void dm_end_request(struct request *clone, int error) | |
761 | { | |
762 | int rw = rq_data_dir(clone); | |
763 | struct dm_rq_target_io *tio = clone->end_io_data; | |
764 | struct mapped_device *md = tio->md; | |
765 | struct request *rq = tio->orig; | |
766 | ||
29e4013d | 767 | if (rq->cmd_type == REQ_TYPE_BLOCK_PC) { |
980691e5 KU |
768 | rq->errors = clone->errors; |
769 | rq->resid_len = clone->resid_len; | |
770 | ||
771 | if (rq->sense) | |
772 | /* | |
773 | * We are using the sense buffer of the original | |
774 | * request. | |
775 | * So setting the length of the sense data is enough. | |
776 | */ | |
777 | rq->sense_len = clone->sense_len; | |
778 | } | |
779 | ||
780 | free_rq_clone(clone); | |
29e4013d TH |
781 | blk_end_request_all(rq, error); |
782 | rq_completed(md, rw, true); | |
980691e5 KU |
783 | } |
784 | ||
cec47e3d KU |
785 | static void dm_unprep_request(struct request *rq) |
786 | { | |
787 | struct request *clone = rq->special; | |
cec47e3d KU |
788 | |
789 | rq->special = NULL; | |
790 | rq->cmd_flags &= ~REQ_DONTPREP; | |
791 | ||
a77e28c7 | 792 | free_rq_clone(clone); |
cec47e3d KU |
793 | } |
794 | ||
795 | /* | |
796 | * Requeue the original request of a clone. | |
797 | */ | |
798 | void dm_requeue_unmapped_request(struct request *clone) | |
799 | { | |
b4324fee | 800 | int rw = rq_data_dir(clone); |
cec47e3d KU |
801 | struct dm_rq_target_io *tio = clone->end_io_data; |
802 | struct mapped_device *md = tio->md; | |
803 | struct request *rq = tio->orig; | |
804 | struct request_queue *q = rq->q; | |
805 | unsigned long flags; | |
806 | ||
807 | dm_unprep_request(rq); | |
808 | ||
809 | spin_lock_irqsave(q->queue_lock, flags); | |
810 | if (elv_queue_empty(q)) | |
811 | blk_plug_device(q); | |
812 | blk_requeue_request(q, rq); | |
813 | spin_unlock_irqrestore(q->queue_lock, flags); | |
814 | ||
b4324fee | 815 | rq_completed(md, rw, 0); |
cec47e3d KU |
816 | } |
817 | EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request); | |
818 | ||
819 | static void __stop_queue(struct request_queue *q) | |
820 | { | |
821 | blk_stop_queue(q); | |
822 | } | |
823 | ||
824 | static void stop_queue(struct request_queue *q) | |
825 | { | |
826 | unsigned long flags; | |
827 | ||
828 | spin_lock_irqsave(q->queue_lock, flags); | |
829 | __stop_queue(q); | |
830 | spin_unlock_irqrestore(q->queue_lock, flags); | |
831 | } | |
832 | ||
833 | static void __start_queue(struct request_queue *q) | |
834 | { | |
835 | if (blk_queue_stopped(q)) | |
836 | blk_start_queue(q); | |
837 | } | |
838 | ||
839 | static void start_queue(struct request_queue *q) | |
840 | { | |
841 | unsigned long flags; | |
842 | ||
843 | spin_lock_irqsave(q->queue_lock, flags); | |
844 | __start_queue(q); | |
845 | spin_unlock_irqrestore(q->queue_lock, flags); | |
846 | } | |
847 | ||
11a68244 | 848 | static void dm_done(struct request *clone, int error, bool mapped) |
cec47e3d | 849 | { |
11a68244 | 850 | int r = error; |
cec47e3d KU |
851 | struct dm_rq_target_io *tio = clone->end_io_data; |
852 | dm_request_endio_fn rq_end_io = tio->ti->type->rq_end_io; | |
cec47e3d | 853 | |
11a68244 KU |
854 | if (mapped && rq_end_io) |
855 | r = rq_end_io(tio->ti, clone, error, &tio->info); | |
cec47e3d | 856 | |
11a68244 | 857 | if (r <= 0) |
cec47e3d | 858 | /* The target wants to complete the I/O */ |
11a68244 KU |
859 | dm_end_request(clone, r); |
860 | else if (r == DM_ENDIO_INCOMPLETE) | |
cec47e3d KU |
861 | /* The target will handle the I/O */ |
862 | return; | |
11a68244 | 863 | else if (r == DM_ENDIO_REQUEUE) |
cec47e3d KU |
864 | /* The target wants to requeue the I/O */ |
865 | dm_requeue_unmapped_request(clone); | |
866 | else { | |
11a68244 | 867 | DMWARN("unimplemented target endio return value: %d", r); |
cec47e3d KU |
868 | BUG(); |
869 | } | |
870 | } | |
871 | ||
11a68244 KU |
872 | /* |
873 | * Request completion handler for request-based dm | |
874 | */ | |
875 | static void dm_softirq_done(struct request *rq) | |
876 | { | |
877 | bool mapped = true; | |
878 | struct request *clone = rq->completion_data; | |
879 | struct dm_rq_target_io *tio = clone->end_io_data; | |
880 | ||
881 | if (rq->cmd_flags & REQ_FAILED) | |
882 | mapped = false; | |
883 | ||
884 | dm_done(clone, tio->error, mapped); | |
885 | } | |
886 | ||
cec47e3d KU |
887 | /* |
888 | * Complete the clone and the original request with the error status | |
889 | * through softirq context. | |
890 | */ | |
891 | static void dm_complete_request(struct request *clone, int error) | |
892 | { | |
893 | struct dm_rq_target_io *tio = clone->end_io_data; | |
894 | struct request *rq = tio->orig; | |
895 | ||
896 | tio->error = error; | |
897 | rq->completion_data = clone; | |
898 | blk_complete_request(rq); | |
899 | } | |
900 | ||
901 | /* | |
902 | * Complete the not-mapped clone and the original request with the error status | |
903 | * through softirq context. | |
904 | * Target's rq_end_io() function isn't called. | |
905 | * This may be used when the target's map_rq() function fails. | |
906 | */ | |
907 | void dm_kill_unmapped_request(struct request *clone, int error) | |
908 | { | |
909 | struct dm_rq_target_io *tio = clone->end_io_data; | |
910 | struct request *rq = tio->orig; | |
911 | ||
912 | rq->cmd_flags |= REQ_FAILED; | |
913 | dm_complete_request(clone, error); | |
914 | } | |
915 | EXPORT_SYMBOL_GPL(dm_kill_unmapped_request); | |
916 | ||
917 | /* | |
918 | * Called with the queue lock held | |
919 | */ | |
920 | static void end_clone_request(struct request *clone, int error) | |
921 | { | |
922 | /* | |
923 | * For just cleaning up the information of the queue in which | |
924 | * the clone was dispatched. | |
925 | * The clone is *NOT* freed actually here because it is alloced from | |
926 | * dm own mempool and REQ_ALLOCED isn't set in clone->cmd_flags. | |
927 | */ | |
928 | __blk_put_request(clone->q, clone); | |
929 | ||
930 | /* | |
931 | * Actual request completion is done in a softirq context which doesn't | |
932 | * hold the queue lock. Otherwise, deadlock could occur because: | |
933 | * - another request may be submitted by the upper level driver | |
934 | * of the stacking during the completion | |
935 | * - the submission which requires queue lock may be done | |
936 | * against this queue | |
937 | */ | |
938 | dm_complete_request(clone, error); | |
939 | } | |
940 | ||
56a67df7 MS |
941 | /* |
942 | * Return maximum size of I/O possible at the supplied sector up to the current | |
943 | * target boundary. | |
944 | */ | |
945 | static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti) | |
946 | { | |
947 | sector_t target_offset = dm_target_offset(ti, sector); | |
948 | ||
949 | return ti->len - target_offset; | |
950 | } | |
951 | ||
952 | static sector_t max_io_len(sector_t sector, struct dm_target *ti) | |
1da177e4 | 953 | { |
56a67df7 | 954 | sector_t len = max_io_len_target_boundary(sector, ti); |
1da177e4 LT |
955 | |
956 | /* | |
957 | * Does the target need to split even further ? | |
958 | */ | |
959 | if (ti->split_io) { | |
960 | sector_t boundary; | |
56a67df7 | 961 | sector_t offset = dm_target_offset(ti, sector); |
1da177e4 LT |
962 | boundary = ((offset + ti->split_io) & ~(ti->split_io - 1)) |
963 | - offset; | |
964 | if (len > boundary) | |
965 | len = boundary; | |
966 | } | |
967 | ||
968 | return len; | |
969 | } | |
970 | ||
971 | static void __map_bio(struct dm_target *ti, struct bio *clone, | |
028867ac | 972 | struct dm_target_io *tio) |
1da177e4 LT |
973 | { |
974 | int r; | |
2056a782 | 975 | sector_t sector; |
9faf400f | 976 | struct mapped_device *md; |
1da177e4 | 977 | |
1da177e4 LT |
978 | clone->bi_end_io = clone_endio; |
979 | clone->bi_private = tio; | |
980 | ||
981 | /* | |
982 | * Map the clone. If r == 0 we don't need to do | |
983 | * anything, the target has assumed ownership of | |
984 | * this io. | |
985 | */ | |
986 | atomic_inc(&tio->io->io_count); | |
2056a782 | 987 | sector = clone->bi_sector; |
1da177e4 | 988 | r = ti->type->map(ti, clone, &tio->info); |
45cbcd79 | 989 | if (r == DM_MAPIO_REMAPPED) { |
1da177e4 | 990 | /* the bio has been remapped so dispatch it */ |
2056a782 | 991 | |
5f3ea37c | 992 | trace_block_remap(bdev_get_queue(clone->bi_bdev), clone, |
22a7c31a | 993 | tio->io->bio->bi_bdev->bd_dev, sector); |
2056a782 | 994 | |
1da177e4 | 995 | generic_make_request(clone); |
2e93ccc1 KU |
996 | } else if (r < 0 || r == DM_MAPIO_REQUEUE) { |
997 | /* error the io and bail out, or requeue it if needed */ | |
9faf400f SB |
998 | md = tio->io->md; |
999 | dec_pending(tio->io, r); | |
1000 | /* | |
1001 | * Store bio_set for cleanup. | |
1002 | */ | |
1003 | clone->bi_private = md->bs; | |
1da177e4 | 1004 | bio_put(clone); |
9faf400f | 1005 | free_tio(md, tio); |
45cbcd79 KU |
1006 | } else if (r) { |
1007 | DMWARN("unimplemented target map return value: %d", r); | |
1008 | BUG(); | |
1da177e4 LT |
1009 | } |
1010 | } | |
1011 | ||
1012 | struct clone_info { | |
1013 | struct mapped_device *md; | |
1014 | struct dm_table *map; | |
1015 | struct bio *bio; | |
1016 | struct dm_io *io; | |
1017 | sector_t sector; | |
1018 | sector_t sector_count; | |
1019 | unsigned short idx; | |
1020 | }; | |
1021 | ||
3676347a PO |
1022 | static void dm_bio_destructor(struct bio *bio) |
1023 | { | |
9faf400f SB |
1024 | struct bio_set *bs = bio->bi_private; |
1025 | ||
1026 | bio_free(bio, bs); | |
3676347a PO |
1027 | } |
1028 | ||
1da177e4 | 1029 | /* |
d87f4c14 | 1030 | * Creates a little bio that just does part of a bvec. |
1da177e4 LT |
1031 | */ |
1032 | static struct bio *split_bvec(struct bio *bio, sector_t sector, | |
1033 | unsigned short idx, unsigned int offset, | |
9faf400f | 1034 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
1035 | { |
1036 | struct bio *clone; | |
1037 | struct bio_vec *bv = bio->bi_io_vec + idx; | |
1038 | ||
9faf400f | 1039 | clone = bio_alloc_bioset(GFP_NOIO, 1, bs); |
3676347a | 1040 | clone->bi_destructor = dm_bio_destructor; |
1da177e4 LT |
1041 | *clone->bi_io_vec = *bv; |
1042 | ||
1043 | clone->bi_sector = sector; | |
1044 | clone->bi_bdev = bio->bi_bdev; | |
d87f4c14 | 1045 | clone->bi_rw = bio->bi_rw; |
1da177e4 LT |
1046 | clone->bi_vcnt = 1; |
1047 | clone->bi_size = to_bytes(len); | |
1048 | clone->bi_io_vec->bv_offset = offset; | |
1049 | clone->bi_io_vec->bv_len = clone->bi_size; | |
f3e1d26e | 1050 | clone->bi_flags |= 1 << BIO_CLONED; |
1da177e4 | 1051 | |
9c47008d | 1052 | if (bio_integrity(bio)) { |
7878cba9 | 1053 | bio_integrity_clone(clone, bio, GFP_NOIO, bs); |
9c47008d MP |
1054 | bio_integrity_trim(clone, |
1055 | bio_sector_offset(bio, idx, offset), len); | |
1056 | } | |
1057 | ||
1da177e4 LT |
1058 | return clone; |
1059 | } | |
1060 | ||
1061 | /* | |
1062 | * Creates a bio that consists of range of complete bvecs. | |
1063 | */ | |
1064 | static struct bio *clone_bio(struct bio *bio, sector_t sector, | |
1065 | unsigned short idx, unsigned short bv_count, | |
9faf400f | 1066 | unsigned int len, struct bio_set *bs) |
1da177e4 LT |
1067 | { |
1068 | struct bio *clone; | |
1069 | ||
9faf400f SB |
1070 | clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs); |
1071 | __bio_clone(clone, bio); | |
1072 | clone->bi_destructor = dm_bio_destructor; | |
1da177e4 LT |
1073 | clone->bi_sector = sector; |
1074 | clone->bi_idx = idx; | |
1075 | clone->bi_vcnt = idx + bv_count; | |
1076 | clone->bi_size = to_bytes(len); | |
1077 | clone->bi_flags &= ~(1 << BIO_SEG_VALID); | |
1078 | ||
9c47008d | 1079 | if (bio_integrity(bio)) { |
7878cba9 | 1080 | bio_integrity_clone(clone, bio, GFP_NOIO, bs); |
9c47008d MP |
1081 | |
1082 | if (idx != bio->bi_idx || clone->bi_size < bio->bi_size) | |
1083 | bio_integrity_trim(clone, | |
1084 | bio_sector_offset(bio, idx, 0), len); | |
1085 | } | |
1086 | ||
1da177e4 LT |
1087 | return clone; |
1088 | } | |
1089 | ||
9015df24 AK |
1090 | static struct dm_target_io *alloc_tio(struct clone_info *ci, |
1091 | struct dm_target *ti) | |
f9ab94ce | 1092 | { |
9015df24 | 1093 | struct dm_target_io *tio = mempool_alloc(ci->md->tio_pool, GFP_NOIO); |
f9ab94ce MP |
1094 | |
1095 | tio->io = ci->io; | |
1096 | tio->ti = ti; | |
f9ab94ce | 1097 | memset(&tio->info, 0, sizeof(tio->info)); |
9015df24 AK |
1098 | |
1099 | return tio; | |
1100 | } | |
1101 | ||
06a426ce | 1102 | static void __issue_target_request(struct clone_info *ci, struct dm_target *ti, |
a79245b3 | 1103 | unsigned request_nr, sector_t len) |
9015df24 AK |
1104 | { |
1105 | struct dm_target_io *tio = alloc_tio(ci, ti); | |
1106 | struct bio *clone; | |
1107 | ||
57cba5d3 | 1108 | tio->info.target_request_nr = request_nr; |
f9ab94ce | 1109 | |
06a426ce MS |
1110 | /* |
1111 | * Discard requests require the bio's inline iovecs be initialized. | |
1112 | * ci->bio->bi_max_vecs is BIO_INLINE_VECS anyway, for both flush | |
1113 | * and discard, so no need for concern about wasted bvec allocations. | |
1114 | */ | |
1115 | clone = bio_alloc_bioset(GFP_NOIO, ci->bio->bi_max_vecs, ci->md->bs); | |
f9ab94ce MP |
1116 | __bio_clone(clone, ci->bio); |
1117 | clone->bi_destructor = dm_bio_destructor; | |
a79245b3 MS |
1118 | if (len) { |
1119 | clone->bi_sector = ci->sector; | |
1120 | clone->bi_size = to_bytes(len); | |
1121 | } | |
f9ab94ce MP |
1122 | |
1123 | __map_bio(ti, clone, tio); | |
1124 | } | |
1125 | ||
06a426ce | 1126 | static void __issue_target_requests(struct clone_info *ci, struct dm_target *ti, |
a79245b3 | 1127 | unsigned num_requests, sector_t len) |
06a426ce MS |
1128 | { |
1129 | unsigned request_nr; | |
1130 | ||
1131 | for (request_nr = 0; request_nr < num_requests; request_nr++) | |
a79245b3 | 1132 | __issue_target_request(ci, ti, request_nr, len); |
06a426ce MS |
1133 | } |
1134 | ||
d87f4c14 | 1135 | static int __clone_and_map_flush(struct clone_info *ci) |
f9ab94ce | 1136 | { |
06a426ce | 1137 | unsigned target_nr = 0; |
f9ab94ce MP |
1138 | struct dm_target *ti; |
1139 | ||
1140 | while ((ti = dm_table_get_target(ci->map, target_nr++))) | |
a79245b3 | 1141 | __issue_target_requests(ci, ti, ti->num_flush_requests, 0); |
f9ab94ce MP |
1142 | |
1143 | ci->sector_count = 0; | |
1144 | ||
1145 | return 0; | |
1146 | } | |
1147 | ||
5ae89a87 MS |
1148 | /* |
1149 | * Perform all io with a single clone. | |
1150 | */ | |
1151 | static void __clone_and_map_simple(struct clone_info *ci, struct dm_target *ti) | |
1152 | { | |
1153 | struct bio *clone, *bio = ci->bio; | |
1154 | struct dm_target_io *tio; | |
1155 | ||
1156 | tio = alloc_tio(ci, ti); | |
1157 | clone = clone_bio(bio, ci->sector, ci->idx, | |
1158 | bio->bi_vcnt - ci->idx, ci->sector_count, | |
1159 | ci->md->bs); | |
1160 | __map_bio(ti, clone, tio); | |
1161 | ci->sector_count = 0; | |
1162 | } | |
1163 | ||
1164 | static int __clone_and_map_discard(struct clone_info *ci) | |
1165 | { | |
1166 | struct dm_target *ti; | |
a79245b3 | 1167 | sector_t len; |
5ae89a87 | 1168 | |
a79245b3 MS |
1169 | do { |
1170 | ti = dm_table_find_target(ci->map, ci->sector); | |
1171 | if (!dm_target_is_valid(ti)) | |
1172 | return -EIO; | |
5ae89a87 | 1173 | |
5ae89a87 | 1174 | /* |
a79245b3 MS |
1175 | * Even though the device advertised discard support, |
1176 | * reconfiguration might have changed that since the | |
1177 | * check was performed. | |
5ae89a87 | 1178 | */ |
a79245b3 MS |
1179 | if (!ti->num_discard_requests) |
1180 | return -EOPNOTSUPP; | |
5ae89a87 | 1181 | |
a79245b3 | 1182 | len = min(ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); |
06a426ce | 1183 | |
a79245b3 MS |
1184 | __issue_target_requests(ci, ti, ti->num_discard_requests, len); |
1185 | ||
1186 | ci->sector += len; | |
1187 | } while (ci->sector_count -= len); | |
5ae89a87 MS |
1188 | |
1189 | return 0; | |
1190 | } | |
1191 | ||
512875bd | 1192 | static int __clone_and_map(struct clone_info *ci) |
1da177e4 LT |
1193 | { |
1194 | struct bio *clone, *bio = ci->bio; | |
512875bd JN |
1195 | struct dm_target *ti; |
1196 | sector_t len = 0, max; | |
028867ac | 1197 | struct dm_target_io *tio; |
1da177e4 | 1198 | |
5ae89a87 MS |
1199 | if (unlikely(bio->bi_rw & REQ_DISCARD)) |
1200 | return __clone_and_map_discard(ci); | |
1201 | ||
512875bd JN |
1202 | ti = dm_table_find_target(ci->map, ci->sector); |
1203 | if (!dm_target_is_valid(ti)) | |
1204 | return -EIO; | |
1205 | ||
56a67df7 | 1206 | max = max_io_len(ci->sector, ti); |
512875bd | 1207 | |
1da177e4 LT |
1208 | if (ci->sector_count <= max) { |
1209 | /* | |
1210 | * Optimise for the simple case where we can do all of | |
1211 | * the remaining io with a single clone. | |
1212 | */ | |
5ae89a87 | 1213 | __clone_and_map_simple(ci, ti); |
1da177e4 LT |
1214 | |
1215 | } else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) { | |
1216 | /* | |
1217 | * There are some bvecs that don't span targets. | |
1218 | * Do as many of these as possible. | |
1219 | */ | |
1220 | int i; | |
1221 | sector_t remaining = max; | |
1222 | sector_t bv_len; | |
1223 | ||
1224 | for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) { | |
1225 | bv_len = to_sector(bio->bi_io_vec[i].bv_len); | |
1226 | ||
1227 | if (bv_len > remaining) | |
1228 | break; | |
1229 | ||
1230 | remaining -= bv_len; | |
1231 | len += bv_len; | |
1232 | } | |
1233 | ||
5ae89a87 | 1234 | tio = alloc_tio(ci, ti); |
9faf400f SB |
1235 | clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len, |
1236 | ci->md->bs); | |
1da177e4 LT |
1237 | __map_bio(ti, clone, tio); |
1238 | ||
1239 | ci->sector += len; | |
1240 | ci->sector_count -= len; | |
1241 | ci->idx = i; | |
1242 | ||
1243 | } else { | |
1244 | /* | |
d2044a94 | 1245 | * Handle a bvec that must be split between two or more targets. |
1da177e4 LT |
1246 | */ |
1247 | struct bio_vec *bv = bio->bi_io_vec + ci->idx; | |
d2044a94 AK |
1248 | sector_t remaining = to_sector(bv->bv_len); |
1249 | unsigned int offset = 0; | |
1da177e4 | 1250 | |
d2044a94 AK |
1251 | do { |
1252 | if (offset) { | |
1253 | ti = dm_table_find_target(ci->map, ci->sector); | |
512875bd JN |
1254 | if (!dm_target_is_valid(ti)) |
1255 | return -EIO; | |
1256 | ||
56a67df7 | 1257 | max = max_io_len(ci->sector, ti); |
d2044a94 AK |
1258 | } |
1259 | ||
1260 | len = min(remaining, max); | |
1261 | ||
5ae89a87 | 1262 | tio = alloc_tio(ci, ti); |
d2044a94 | 1263 | clone = split_bvec(bio, ci->sector, ci->idx, |
9faf400f SB |
1264 | bv->bv_offset + offset, len, |
1265 | ci->md->bs); | |
d2044a94 AK |
1266 | |
1267 | __map_bio(ti, clone, tio); | |
1268 | ||
1269 | ci->sector += len; | |
1270 | ci->sector_count -= len; | |
1271 | offset += to_bytes(len); | |
1272 | } while (remaining -= len); | |
1da177e4 | 1273 | |
1da177e4 LT |
1274 | ci->idx++; |
1275 | } | |
512875bd JN |
1276 | |
1277 | return 0; | |
1da177e4 LT |
1278 | } |
1279 | ||
1280 | /* | |
8a53c28d | 1281 | * Split the bio into several clones and submit it to targets. |
1da177e4 | 1282 | */ |
f0b9a450 | 1283 | static void __split_and_process_bio(struct mapped_device *md, struct bio *bio) |
1da177e4 | 1284 | { |
6a8736d1 | 1285 | bool is_flush = bio->bi_rw & REQ_FLUSH; |
1da177e4 | 1286 | struct clone_info ci; |
512875bd | 1287 | int error = 0; |
1da177e4 | 1288 | |
7c666411 | 1289 | ci.map = dm_get_live_table(md); |
f0b9a450 | 1290 | if (unlikely(!ci.map)) { |
6a8736d1 | 1291 | bio_io_error(bio); |
f0b9a450 MP |
1292 | return; |
1293 | } | |
692d0eb9 | 1294 | |
1da177e4 | 1295 | ci.md = md; |
1da177e4 LT |
1296 | ci.io = alloc_io(md); |
1297 | ci.io->error = 0; | |
1298 | atomic_set(&ci.io->io_count, 1); | |
1299 | ci.io->bio = bio; | |
1300 | ci.io->md = md; | |
f88fb981 | 1301 | spin_lock_init(&ci.io->endio_lock); |
1da177e4 | 1302 | ci.sector = bio->bi_sector; |
6a8736d1 TH |
1303 | ci.idx = bio->bi_idx; |
1304 | ||
1305 | if (!is_flush) { | |
1306 | ci.bio = bio; | |
d87f4c14 | 1307 | ci.sector_count = bio_sectors(bio); |
6a8736d1 TH |
1308 | } else { |
1309 | ci.bio = &ci.md->flush_bio; | |
f9ab94ce | 1310 | ci.sector_count = 1; |
d87f4c14 | 1311 | } |
1da177e4 | 1312 | |
3eaf840e | 1313 | start_io_acct(ci.io); |
d87f4c14 | 1314 | while (ci.sector_count && !error) { |
6a8736d1 | 1315 | if (!is_flush) |
d87f4c14 TH |
1316 | error = __clone_and_map(&ci); |
1317 | else | |
1318 | error = __clone_and_map_flush(&ci); | |
1319 | } | |
1da177e4 LT |
1320 | |
1321 | /* drop the extra reference count */ | |
512875bd | 1322 | dec_pending(ci.io, error); |
1da177e4 LT |
1323 | dm_table_put(ci.map); |
1324 | } | |
1325 | /*----------------------------------------------------------------- | |
1326 | * CRUD END | |
1327 | *---------------------------------------------------------------*/ | |
1328 | ||
f6fccb12 MB |
1329 | static int dm_merge_bvec(struct request_queue *q, |
1330 | struct bvec_merge_data *bvm, | |
1331 | struct bio_vec *biovec) | |
1332 | { | |
1333 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1334 | struct dm_table *map = dm_get_live_table(md); |
f6fccb12 MB |
1335 | struct dm_target *ti; |
1336 | sector_t max_sectors; | |
5037108a | 1337 | int max_size = 0; |
f6fccb12 MB |
1338 | |
1339 | if (unlikely(!map)) | |
5037108a | 1340 | goto out; |
f6fccb12 MB |
1341 | |
1342 | ti = dm_table_find_target(map, bvm->bi_sector); | |
b01cd5ac MP |
1343 | if (!dm_target_is_valid(ti)) |
1344 | goto out_table; | |
f6fccb12 MB |
1345 | |
1346 | /* | |
1347 | * Find maximum amount of I/O that won't need splitting | |
1348 | */ | |
56a67df7 | 1349 | max_sectors = min(max_io_len(bvm->bi_sector, ti), |
f6fccb12 MB |
1350 | (sector_t) BIO_MAX_SECTORS); |
1351 | max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size; | |
1352 | if (max_size < 0) | |
1353 | max_size = 0; | |
1354 | ||
1355 | /* | |
1356 | * merge_bvec_fn() returns number of bytes | |
1357 | * it can accept at this offset | |
1358 | * max is precomputed maximal io size | |
1359 | */ | |
1360 | if (max_size && ti->type->merge) | |
1361 | max_size = ti->type->merge(ti, bvm, biovec, max_size); | |
8cbeb67a MP |
1362 | /* |
1363 | * If the target doesn't support merge method and some of the devices | |
1364 | * provided their merge_bvec method (we know this by looking at | |
1365 | * queue_max_hw_sectors), then we can't allow bios with multiple vector | |
1366 | * entries. So always set max_size to 0, and the code below allows | |
1367 | * just one page. | |
1368 | */ | |
1369 | else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9) | |
1370 | ||
1371 | max_size = 0; | |
f6fccb12 | 1372 | |
b01cd5ac | 1373 | out_table: |
5037108a MP |
1374 | dm_table_put(map); |
1375 | ||
1376 | out: | |
f6fccb12 MB |
1377 | /* |
1378 | * Always allow an entire first page | |
1379 | */ | |
1380 | if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT)) | |
1381 | max_size = biovec->bv_len; | |
1382 | ||
f6fccb12 MB |
1383 | return max_size; |
1384 | } | |
1385 | ||
1da177e4 LT |
1386 | /* |
1387 | * The request function that just remaps the bio built up by | |
1388 | * dm_merge_bvec. | |
1389 | */ | |
cec47e3d | 1390 | static int _dm_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 1391 | { |
12f03a49 | 1392 | int rw = bio_data_dir(bio); |
1da177e4 | 1393 | struct mapped_device *md = q->queuedata; |
c9959059 | 1394 | int cpu; |
1da177e4 | 1395 | |
2ca3310e | 1396 | down_read(&md->io_lock); |
1da177e4 | 1397 | |
074a7aca TH |
1398 | cpu = part_stat_lock(); |
1399 | part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]); | |
1400 | part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio)); | |
1401 | part_stat_unlock(); | |
12f03a49 | 1402 | |
6a8736d1 TH |
1403 | /* if we're suspended, we have to queue this io for later */ |
1404 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) { | |
2ca3310e | 1405 | up_read(&md->io_lock); |
1da177e4 | 1406 | |
6a8736d1 TH |
1407 | if (bio_rw(bio) != READA) |
1408 | queue_io(md, bio); | |
1409 | else | |
54d9a1b4 | 1410 | bio_io_error(bio); |
92c63902 | 1411 | return 0; |
1da177e4 LT |
1412 | } |
1413 | ||
f0b9a450 | 1414 | __split_and_process_bio(md, bio); |
2ca3310e | 1415 | up_read(&md->io_lock); |
f0b9a450 | 1416 | return 0; |
1da177e4 LT |
1417 | } |
1418 | ||
cec47e3d KU |
1419 | static int dm_make_request(struct request_queue *q, struct bio *bio) |
1420 | { | |
1421 | struct mapped_device *md = q->queuedata; | |
1422 | ||
cec47e3d KU |
1423 | return md->saved_make_request_fn(q, bio); /* call __make_request() */ |
1424 | } | |
1425 | ||
1426 | static int dm_request_based(struct mapped_device *md) | |
1427 | { | |
1428 | return blk_queue_stackable(md->queue); | |
1429 | } | |
1430 | ||
1431 | static int dm_request(struct request_queue *q, struct bio *bio) | |
1432 | { | |
1433 | struct mapped_device *md = q->queuedata; | |
1434 | ||
1435 | if (dm_request_based(md)) | |
1436 | return dm_make_request(q, bio); | |
1437 | ||
1438 | return _dm_request(q, bio); | |
1439 | } | |
1440 | ||
1441 | void dm_dispatch_request(struct request *rq) | |
1442 | { | |
1443 | int r; | |
1444 | ||
1445 | if (blk_queue_io_stat(rq->q)) | |
1446 | rq->cmd_flags |= REQ_IO_STAT; | |
1447 | ||
1448 | rq->start_time = jiffies; | |
1449 | r = blk_insert_cloned_request(rq->q, rq); | |
1450 | if (r) | |
1451 | dm_complete_request(rq, r); | |
1452 | } | |
1453 | EXPORT_SYMBOL_GPL(dm_dispatch_request); | |
1454 | ||
1455 | static void dm_rq_bio_destructor(struct bio *bio) | |
1456 | { | |
1457 | struct dm_rq_clone_bio_info *info = bio->bi_private; | |
1458 | struct mapped_device *md = info->tio->md; | |
1459 | ||
1460 | free_bio_info(info); | |
1461 | bio_free(bio, md->bs); | |
1462 | } | |
1463 | ||
1464 | static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig, | |
1465 | void *data) | |
1466 | { | |
1467 | struct dm_rq_target_io *tio = data; | |
1468 | struct mapped_device *md = tio->md; | |
1469 | struct dm_rq_clone_bio_info *info = alloc_bio_info(md); | |
1470 | ||
1471 | if (!info) | |
1472 | return -ENOMEM; | |
1473 | ||
1474 | info->orig = bio_orig; | |
1475 | info->tio = tio; | |
1476 | bio->bi_end_io = end_clone_bio; | |
1477 | bio->bi_private = info; | |
1478 | bio->bi_destructor = dm_rq_bio_destructor; | |
1479 | ||
1480 | return 0; | |
1481 | } | |
1482 | ||
1483 | static int setup_clone(struct request *clone, struct request *rq, | |
1484 | struct dm_rq_target_io *tio) | |
1485 | { | |
d0bcb878 | 1486 | int r; |
cec47e3d | 1487 | |
29e4013d TH |
1488 | r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC, |
1489 | dm_rq_bio_constructor, tio); | |
1490 | if (r) | |
1491 | return r; | |
cec47e3d | 1492 | |
29e4013d TH |
1493 | clone->cmd = rq->cmd; |
1494 | clone->cmd_len = rq->cmd_len; | |
1495 | clone->sense = rq->sense; | |
1496 | clone->buffer = rq->buffer; | |
cec47e3d KU |
1497 | clone->end_io = end_clone_request; |
1498 | clone->end_io_data = tio; | |
1499 | ||
1500 | return 0; | |
1501 | } | |
1502 | ||
6facdaff KU |
1503 | static struct request *clone_rq(struct request *rq, struct mapped_device *md, |
1504 | gfp_t gfp_mask) | |
1505 | { | |
1506 | struct request *clone; | |
1507 | struct dm_rq_target_io *tio; | |
1508 | ||
1509 | tio = alloc_rq_tio(md, gfp_mask); | |
1510 | if (!tio) | |
1511 | return NULL; | |
1512 | ||
1513 | tio->md = md; | |
1514 | tio->ti = NULL; | |
1515 | tio->orig = rq; | |
1516 | tio->error = 0; | |
1517 | memset(&tio->info, 0, sizeof(tio->info)); | |
1518 | ||
1519 | clone = &tio->clone; | |
1520 | if (setup_clone(clone, rq, tio)) { | |
1521 | /* -ENOMEM */ | |
1522 | free_rq_tio(tio); | |
1523 | return NULL; | |
1524 | } | |
1525 | ||
1526 | return clone; | |
1527 | } | |
1528 | ||
cec47e3d KU |
1529 | /* |
1530 | * Called with the queue lock held. | |
1531 | */ | |
1532 | static int dm_prep_fn(struct request_queue *q, struct request *rq) | |
1533 | { | |
1534 | struct mapped_device *md = q->queuedata; | |
cec47e3d KU |
1535 | struct request *clone; |
1536 | ||
cec47e3d KU |
1537 | if (unlikely(rq->special)) { |
1538 | DMWARN("Already has something in rq->special."); | |
1539 | return BLKPREP_KILL; | |
1540 | } | |
1541 | ||
6facdaff KU |
1542 | clone = clone_rq(rq, md, GFP_ATOMIC); |
1543 | if (!clone) | |
cec47e3d | 1544 | return BLKPREP_DEFER; |
cec47e3d KU |
1545 | |
1546 | rq->special = clone; | |
1547 | rq->cmd_flags |= REQ_DONTPREP; | |
1548 | ||
1549 | return BLKPREP_OK; | |
1550 | } | |
1551 | ||
9eef87da KU |
1552 | /* |
1553 | * Returns: | |
1554 | * 0 : the request has been processed (not requeued) | |
1555 | * !0 : the request has been requeued | |
1556 | */ | |
1557 | static int map_request(struct dm_target *ti, struct request *clone, | |
1558 | struct mapped_device *md) | |
cec47e3d | 1559 | { |
9eef87da | 1560 | int r, requeued = 0; |
cec47e3d KU |
1561 | struct dm_rq_target_io *tio = clone->end_io_data; |
1562 | ||
1563 | /* | |
1564 | * Hold the md reference here for the in-flight I/O. | |
1565 | * We can't rely on the reference count by device opener, | |
1566 | * because the device may be closed during the request completion | |
1567 | * when all bios are completed. | |
1568 | * See the comment in rq_completed() too. | |
1569 | */ | |
1570 | dm_get(md); | |
1571 | ||
1572 | tio->ti = ti; | |
1573 | r = ti->type->map_rq(ti, clone, &tio->info); | |
1574 | switch (r) { | |
1575 | case DM_MAPIO_SUBMITTED: | |
1576 | /* The target has taken the I/O to submit by itself later */ | |
1577 | break; | |
1578 | case DM_MAPIO_REMAPPED: | |
1579 | /* The target has remapped the I/O so dispatch it */ | |
6db4ccd6 JN |
1580 | trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)), |
1581 | blk_rq_pos(tio->orig)); | |
cec47e3d KU |
1582 | dm_dispatch_request(clone); |
1583 | break; | |
1584 | case DM_MAPIO_REQUEUE: | |
1585 | /* The target wants to requeue the I/O */ | |
1586 | dm_requeue_unmapped_request(clone); | |
9eef87da | 1587 | requeued = 1; |
cec47e3d KU |
1588 | break; |
1589 | default: | |
1590 | if (r > 0) { | |
1591 | DMWARN("unimplemented target map return value: %d", r); | |
1592 | BUG(); | |
1593 | } | |
1594 | ||
1595 | /* The target wants to complete the I/O */ | |
1596 | dm_kill_unmapped_request(clone, r); | |
1597 | break; | |
1598 | } | |
9eef87da KU |
1599 | |
1600 | return requeued; | |
cec47e3d KU |
1601 | } |
1602 | ||
1603 | /* | |
1604 | * q->request_fn for request-based dm. | |
1605 | * Called with the queue lock held. | |
1606 | */ | |
1607 | static void dm_request_fn(struct request_queue *q) | |
1608 | { | |
1609 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1610 | struct dm_table *map = dm_get_live_table(md); |
cec47e3d | 1611 | struct dm_target *ti; |
b4324fee | 1612 | struct request *rq, *clone; |
29e4013d | 1613 | sector_t pos; |
cec47e3d KU |
1614 | |
1615 | /* | |
b4324fee KU |
1616 | * For suspend, check blk_queue_stopped() and increment |
1617 | * ->pending within a single queue_lock not to increment the | |
1618 | * number of in-flight I/Os after the queue is stopped in | |
1619 | * dm_suspend(). | |
cec47e3d KU |
1620 | */ |
1621 | while (!blk_queue_plugged(q) && !blk_queue_stopped(q)) { | |
1622 | rq = blk_peek_request(q); | |
1623 | if (!rq) | |
1624 | goto plug_and_out; | |
1625 | ||
29e4013d TH |
1626 | /* always use block 0 to find the target for flushes for now */ |
1627 | pos = 0; | |
1628 | if (!(rq->cmd_flags & REQ_FLUSH)) | |
1629 | pos = blk_rq_pos(rq); | |
1630 | ||
1631 | ti = dm_table_find_target(map, pos); | |
1632 | BUG_ON(!dm_target_is_valid(ti)); | |
d0bcb878 | 1633 | |
cec47e3d KU |
1634 | if (ti->type->busy && ti->type->busy(ti)) |
1635 | goto plug_and_out; | |
1636 | ||
1637 | blk_start_request(rq); | |
b4324fee KU |
1638 | clone = rq->special; |
1639 | atomic_inc(&md->pending[rq_data_dir(clone)]); | |
1640 | ||
cec47e3d | 1641 | spin_unlock(q->queue_lock); |
9eef87da KU |
1642 | if (map_request(ti, clone, md)) |
1643 | goto requeued; | |
1644 | ||
cec47e3d KU |
1645 | spin_lock_irq(q->queue_lock); |
1646 | } | |
1647 | ||
1648 | goto out; | |
1649 | ||
9eef87da KU |
1650 | requeued: |
1651 | spin_lock_irq(q->queue_lock); | |
1652 | ||
cec47e3d KU |
1653 | plug_and_out: |
1654 | if (!elv_queue_empty(q)) | |
1655 | /* Some requests still remain, retry later */ | |
1656 | blk_plug_device(q); | |
1657 | ||
1658 | out: | |
1659 | dm_table_put(map); | |
1660 | ||
1661 | return; | |
1662 | } | |
1663 | ||
1664 | int dm_underlying_device_busy(struct request_queue *q) | |
1665 | { | |
1666 | return blk_lld_busy(q); | |
1667 | } | |
1668 | EXPORT_SYMBOL_GPL(dm_underlying_device_busy); | |
1669 | ||
1670 | static int dm_lld_busy(struct request_queue *q) | |
1671 | { | |
1672 | int r; | |
1673 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1674 | struct dm_table *map = dm_get_live_table(md); |
cec47e3d KU |
1675 | |
1676 | if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) | |
1677 | r = 1; | |
1678 | else | |
1679 | r = dm_table_any_busy_target(map); | |
1680 | ||
1681 | dm_table_put(map); | |
1682 | ||
1683 | return r; | |
1684 | } | |
1685 | ||
165125e1 | 1686 | static void dm_unplug_all(struct request_queue *q) |
1da177e4 LT |
1687 | { |
1688 | struct mapped_device *md = q->queuedata; | |
7c666411 | 1689 | struct dm_table *map = dm_get_live_table(md); |
1da177e4 LT |
1690 | |
1691 | if (map) { | |
cec47e3d KU |
1692 | if (dm_request_based(md)) |
1693 | generic_unplug_device(q); | |
1694 | ||
1da177e4 LT |
1695 | dm_table_unplug_all(map); |
1696 | dm_table_put(map); | |
1697 | } | |
1698 | } | |
1699 | ||
1700 | static int dm_any_congested(void *congested_data, int bdi_bits) | |
1701 | { | |
8a57dfc6 CS |
1702 | int r = bdi_bits; |
1703 | struct mapped_device *md = congested_data; | |
1704 | struct dm_table *map; | |
1da177e4 | 1705 | |
1eb787ec | 1706 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
7c666411 | 1707 | map = dm_get_live_table(md); |
8a57dfc6 | 1708 | if (map) { |
cec47e3d KU |
1709 | /* |
1710 | * Request-based dm cares about only own queue for | |
1711 | * the query about congestion status of request_queue | |
1712 | */ | |
1713 | if (dm_request_based(md)) | |
1714 | r = md->queue->backing_dev_info.state & | |
1715 | bdi_bits; | |
1716 | else | |
1717 | r = dm_table_any_congested(map, bdi_bits); | |
1718 | ||
8a57dfc6 CS |
1719 | dm_table_put(map); |
1720 | } | |
1721 | } | |
1722 | ||
1da177e4 LT |
1723 | return r; |
1724 | } | |
1725 | ||
1726 | /*----------------------------------------------------------------- | |
1727 | * An IDR is used to keep track of allocated minor numbers. | |
1728 | *---------------------------------------------------------------*/ | |
1da177e4 LT |
1729 | static DEFINE_IDR(_minor_idr); |
1730 | ||
2b06cfff | 1731 | static void free_minor(int minor) |
1da177e4 | 1732 | { |
f32c10b0 | 1733 | spin_lock(&_minor_lock); |
1da177e4 | 1734 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1735 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1736 | } |
1737 | ||
1738 | /* | |
1739 | * See if the device with a specific minor # is free. | |
1740 | */ | |
cf13ab8e | 1741 | static int specific_minor(int minor) |
1da177e4 LT |
1742 | { |
1743 | int r, m; | |
1744 | ||
1745 | if (minor >= (1 << MINORBITS)) | |
1746 | return -EINVAL; | |
1747 | ||
62f75c2f JM |
1748 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
1749 | if (!r) | |
1750 | return -ENOMEM; | |
1751 | ||
f32c10b0 | 1752 | spin_lock(&_minor_lock); |
1da177e4 LT |
1753 | |
1754 | if (idr_find(&_minor_idr, minor)) { | |
1755 | r = -EBUSY; | |
1756 | goto out; | |
1757 | } | |
1758 | ||
ba61fdd1 | 1759 | r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m); |
62f75c2f | 1760 | if (r) |
1da177e4 | 1761 | goto out; |
1da177e4 LT |
1762 | |
1763 | if (m != minor) { | |
1764 | idr_remove(&_minor_idr, m); | |
1765 | r = -EBUSY; | |
1766 | goto out; | |
1767 | } | |
1768 | ||
1769 | out: | |
f32c10b0 | 1770 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1771 | return r; |
1772 | } | |
1773 | ||
cf13ab8e | 1774 | static int next_free_minor(int *minor) |
1da177e4 | 1775 | { |
2b06cfff | 1776 | int r, m; |
1da177e4 | 1777 | |
1da177e4 | 1778 | r = idr_pre_get(&_minor_idr, GFP_KERNEL); |
62f75c2f JM |
1779 | if (!r) |
1780 | return -ENOMEM; | |
1781 | ||
f32c10b0 | 1782 | spin_lock(&_minor_lock); |
1da177e4 | 1783 | |
ba61fdd1 | 1784 | r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m); |
cf13ab8e | 1785 | if (r) |
1da177e4 | 1786 | goto out; |
1da177e4 LT |
1787 | |
1788 | if (m >= (1 << MINORBITS)) { | |
1789 | idr_remove(&_minor_idr, m); | |
1790 | r = -ENOSPC; | |
1791 | goto out; | |
1792 | } | |
1793 | ||
1794 | *minor = m; | |
1795 | ||
1796 | out: | |
f32c10b0 | 1797 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1798 | return r; |
1799 | } | |
1800 | ||
83d5cde4 | 1801 | static const struct block_device_operations dm_blk_dops; |
1da177e4 | 1802 | |
53d5914f MP |
1803 | static void dm_wq_work(struct work_struct *work); |
1804 | ||
4a0b4ddf MS |
1805 | static void dm_init_md_queue(struct mapped_device *md) |
1806 | { | |
1807 | /* | |
1808 | * Request-based dm devices cannot be stacked on top of bio-based dm | |
1809 | * devices. The type of this dm device has not been decided yet. | |
1810 | * The type is decided at the first table loading time. | |
1811 | * To prevent problematic device stacking, clear the queue flag | |
1812 | * for request stacking support until then. | |
1813 | * | |
1814 | * This queue is new, so no concurrency on the queue_flags. | |
1815 | */ | |
1816 | queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue); | |
1817 | ||
1818 | md->queue->queuedata = md; | |
1819 | md->queue->backing_dev_info.congested_fn = dm_any_congested; | |
1820 | md->queue->backing_dev_info.congested_data = md; | |
1821 | blk_queue_make_request(md->queue, dm_request); | |
1822 | blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY); | |
1823 | md->queue->unplug_fn = dm_unplug_all; | |
1824 | blk_queue_merge_bvec(md->queue, dm_merge_bvec); | |
d87f4c14 | 1825 | blk_queue_flush(md->queue, REQ_FLUSH | REQ_FUA); |
4a0b4ddf MS |
1826 | } |
1827 | ||
1da177e4 LT |
1828 | /* |
1829 | * Allocate and initialise a blank device with a given minor. | |
1830 | */ | |
2b06cfff | 1831 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 LT |
1832 | { |
1833 | int r; | |
cf13ab8e | 1834 | struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL); |
ba61fdd1 | 1835 | void *old_md; |
1da177e4 LT |
1836 | |
1837 | if (!md) { | |
1838 | DMWARN("unable to allocate device, out of memory."); | |
1839 | return NULL; | |
1840 | } | |
1841 | ||
10da4f79 | 1842 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1843 | goto bad_module_get; |
10da4f79 | 1844 | |
1da177e4 | 1845 | /* get a minor number for the dev */ |
2b06cfff | 1846 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1847 | r = next_free_minor(&minor); |
2b06cfff | 1848 | else |
cf13ab8e | 1849 | r = specific_minor(minor); |
1da177e4 | 1850 | if (r < 0) |
6ed7ade8 | 1851 | goto bad_minor; |
1da177e4 | 1852 | |
a5664dad | 1853 | md->type = DM_TYPE_NONE; |
2ca3310e | 1854 | init_rwsem(&md->io_lock); |
e61290a4 | 1855 | mutex_init(&md->suspend_lock); |
a5664dad | 1856 | mutex_init(&md->type_lock); |
022c2611 | 1857 | spin_lock_init(&md->deferred_lock); |
1da177e4 LT |
1858 | rwlock_init(&md->map_lock); |
1859 | atomic_set(&md->holders, 1); | |
5c6bd75d | 1860 | atomic_set(&md->open_count, 0); |
1da177e4 | 1861 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1862 | atomic_set(&md->uevent_seq, 0); |
1863 | INIT_LIST_HEAD(&md->uevent_list); | |
1864 | spin_lock_init(&md->uevent_lock); | |
1da177e4 | 1865 | |
4a0b4ddf | 1866 | md->queue = blk_alloc_queue(GFP_KERNEL); |
1da177e4 | 1867 | if (!md->queue) |
6ed7ade8 | 1868 | goto bad_queue; |
1da177e4 | 1869 | |
4a0b4ddf | 1870 | dm_init_md_queue(md); |
9faf400f | 1871 | |
1da177e4 LT |
1872 | md->disk = alloc_disk(1); |
1873 | if (!md->disk) | |
6ed7ade8 | 1874 | goto bad_disk; |
1da177e4 | 1875 | |
316d315b NK |
1876 | atomic_set(&md->pending[0], 0); |
1877 | atomic_set(&md->pending[1], 0); | |
f0b04115 | 1878 | init_waitqueue_head(&md->wait); |
53d5914f | 1879 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 JM |
1880 | init_waitqueue_head(&md->eventq); |
1881 | ||
1da177e4 LT |
1882 | md->disk->major = _major; |
1883 | md->disk->first_minor = minor; | |
1884 | md->disk->fops = &dm_blk_dops; | |
1885 | md->disk->queue = md->queue; | |
1886 | md->disk->private_data = md; | |
1887 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
1888 | add_disk(md->disk); | |
7e51f257 | 1889 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1890 | |
304f3f6a MB |
1891 | md->wq = create_singlethread_workqueue("kdmflush"); |
1892 | if (!md->wq) | |
1893 | goto bad_thread; | |
1894 | ||
32a926da MP |
1895 | md->bdev = bdget_disk(md->disk, 0); |
1896 | if (!md->bdev) | |
1897 | goto bad_bdev; | |
1898 | ||
6a8736d1 TH |
1899 | bio_init(&md->flush_bio); |
1900 | md->flush_bio.bi_bdev = md->bdev; | |
1901 | md->flush_bio.bi_rw = WRITE_FLUSH; | |
1902 | ||
ba61fdd1 | 1903 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1904 | spin_lock(&_minor_lock); |
ba61fdd1 | 1905 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1906 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1907 | |
1908 | BUG_ON(old_md != MINOR_ALLOCED); | |
1909 | ||
1da177e4 LT |
1910 | return md; |
1911 | ||
32a926da MP |
1912 | bad_bdev: |
1913 | destroy_workqueue(md->wq); | |
304f3f6a | 1914 | bad_thread: |
03022c54 | 1915 | del_gendisk(md->disk); |
304f3f6a | 1916 | put_disk(md->disk); |
6ed7ade8 | 1917 | bad_disk: |
1312f40e | 1918 | blk_cleanup_queue(md->queue); |
6ed7ade8 | 1919 | bad_queue: |
1da177e4 | 1920 | free_minor(minor); |
6ed7ade8 | 1921 | bad_minor: |
10da4f79 | 1922 | module_put(THIS_MODULE); |
6ed7ade8 | 1923 | bad_module_get: |
1da177e4 LT |
1924 | kfree(md); |
1925 | return NULL; | |
1926 | } | |
1927 | ||
ae9da83f JN |
1928 | static void unlock_fs(struct mapped_device *md); |
1929 | ||
1da177e4 LT |
1930 | static void free_dev(struct mapped_device *md) |
1931 | { | |
f331c029 | 1932 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1933 | |
32a926da MP |
1934 | unlock_fs(md); |
1935 | bdput(md->bdev); | |
304f3f6a | 1936 | destroy_workqueue(md->wq); |
e6ee8c0b KU |
1937 | if (md->tio_pool) |
1938 | mempool_destroy(md->tio_pool); | |
1939 | if (md->io_pool) | |
1940 | mempool_destroy(md->io_pool); | |
1941 | if (md->bs) | |
1942 | bioset_free(md->bs); | |
9c47008d | 1943 | blk_integrity_unregister(md->disk); |
1da177e4 | 1944 | del_gendisk(md->disk); |
63d94e48 | 1945 | free_minor(minor); |
fba9f90e JM |
1946 | |
1947 | spin_lock(&_minor_lock); | |
1948 | md->disk->private_data = NULL; | |
1949 | spin_unlock(&_minor_lock); | |
1950 | ||
1da177e4 | 1951 | put_disk(md->disk); |
1312f40e | 1952 | blk_cleanup_queue(md->queue); |
10da4f79 | 1953 | module_put(THIS_MODULE); |
1da177e4 LT |
1954 | kfree(md); |
1955 | } | |
1956 | ||
e6ee8c0b KU |
1957 | static void __bind_mempools(struct mapped_device *md, struct dm_table *t) |
1958 | { | |
1959 | struct dm_md_mempools *p; | |
1960 | ||
1961 | if (md->io_pool && md->tio_pool && md->bs) | |
1962 | /* the md already has necessary mempools */ | |
1963 | goto out; | |
1964 | ||
1965 | p = dm_table_get_md_mempools(t); | |
1966 | BUG_ON(!p || md->io_pool || md->tio_pool || md->bs); | |
1967 | ||
1968 | md->io_pool = p->io_pool; | |
1969 | p->io_pool = NULL; | |
1970 | md->tio_pool = p->tio_pool; | |
1971 | p->tio_pool = NULL; | |
1972 | md->bs = p->bs; | |
1973 | p->bs = NULL; | |
1974 | ||
1975 | out: | |
1976 | /* mempool bind completed, now no need any mempools in the table */ | |
1977 | dm_table_free_md_mempools(t); | |
1978 | } | |
1979 | ||
1da177e4 LT |
1980 | /* |
1981 | * Bind a table to the device. | |
1982 | */ | |
1983 | static void event_callback(void *context) | |
1984 | { | |
7a8c3d3b MA |
1985 | unsigned long flags; |
1986 | LIST_HEAD(uevents); | |
1da177e4 LT |
1987 | struct mapped_device *md = (struct mapped_device *) context; |
1988 | ||
7a8c3d3b MA |
1989 | spin_lock_irqsave(&md->uevent_lock, flags); |
1990 | list_splice_init(&md->uevent_list, &uevents); | |
1991 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1992 | ||
ed9e1982 | 1993 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1994 | |
1da177e4 LT |
1995 | atomic_inc(&md->event_nr); |
1996 | wake_up(&md->eventq); | |
1997 | } | |
1998 | ||
4e90188b | 1999 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 2000 | { |
4e90188b | 2001 | set_capacity(md->disk, size); |
1da177e4 | 2002 | |
db8fef4f MP |
2003 | mutex_lock(&md->bdev->bd_inode->i_mutex); |
2004 | i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); | |
2005 | mutex_unlock(&md->bdev->bd_inode->i_mutex); | |
1da177e4 LT |
2006 | } |
2007 | ||
042d2a9b AK |
2008 | /* |
2009 | * Returns old map, which caller must destroy. | |
2010 | */ | |
2011 | static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, | |
2012 | struct queue_limits *limits) | |
1da177e4 | 2013 | { |
042d2a9b | 2014 | struct dm_table *old_map; |
165125e1 | 2015 | struct request_queue *q = md->queue; |
1da177e4 | 2016 | sector_t size; |
523d9297 | 2017 | unsigned long flags; |
1da177e4 LT |
2018 | |
2019 | size = dm_table_get_size(t); | |
3ac51e74 DW |
2020 | |
2021 | /* | |
2022 | * Wipe any geometry if the size of the table changed. | |
2023 | */ | |
2024 | if (size != get_capacity(md->disk)) | |
2025 | memset(&md->geometry, 0, sizeof(md->geometry)); | |
2026 | ||
32a926da | 2027 | __set_size(md, size); |
d5816876 | 2028 | |
2ca3310e AK |
2029 | dm_table_event_callback(t, event_callback, md); |
2030 | ||
e6ee8c0b KU |
2031 | /* |
2032 | * The queue hasn't been stopped yet, if the old table type wasn't | |
2033 | * for request-based during suspension. So stop it to prevent | |
2034 | * I/O mapping before resume. | |
2035 | * This must be done before setting the queue restrictions, | |
2036 | * because request-based dm may be run just after the setting. | |
2037 | */ | |
2038 | if (dm_table_request_based(t) && !blk_queue_stopped(q)) | |
2039 | stop_queue(q); | |
2040 | ||
2041 | __bind_mempools(md, t); | |
2042 | ||
523d9297 | 2043 | write_lock_irqsave(&md->map_lock, flags); |
042d2a9b | 2044 | old_map = md->map; |
1da177e4 | 2045 | md->map = t; |
754c5fc7 | 2046 | dm_table_set_restrictions(t, q, limits); |
523d9297 | 2047 | write_unlock_irqrestore(&md->map_lock, flags); |
1da177e4 | 2048 | |
042d2a9b | 2049 | return old_map; |
1da177e4 LT |
2050 | } |
2051 | ||
a7940155 AK |
2052 | /* |
2053 | * Returns unbound table for the caller to free. | |
2054 | */ | |
2055 | static struct dm_table *__unbind(struct mapped_device *md) | |
1da177e4 LT |
2056 | { |
2057 | struct dm_table *map = md->map; | |
523d9297 | 2058 | unsigned long flags; |
1da177e4 LT |
2059 | |
2060 | if (!map) | |
a7940155 | 2061 | return NULL; |
1da177e4 LT |
2062 | |
2063 | dm_table_event_callback(map, NULL, NULL); | |
523d9297 | 2064 | write_lock_irqsave(&md->map_lock, flags); |
1da177e4 | 2065 | md->map = NULL; |
523d9297 | 2066 | write_unlock_irqrestore(&md->map_lock, flags); |
a7940155 AK |
2067 | |
2068 | return map; | |
1da177e4 LT |
2069 | } |
2070 | ||
2071 | /* | |
2072 | * Constructor for a new device. | |
2073 | */ | |
2b06cfff | 2074 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
2075 | { |
2076 | struct mapped_device *md; | |
2077 | ||
2b06cfff | 2078 | md = alloc_dev(minor); |
1da177e4 LT |
2079 | if (!md) |
2080 | return -ENXIO; | |
2081 | ||
784aae73 MB |
2082 | dm_sysfs_init(md); |
2083 | ||
1da177e4 LT |
2084 | *result = md; |
2085 | return 0; | |
2086 | } | |
2087 | ||
a5664dad MS |
2088 | /* |
2089 | * Functions to manage md->type. | |
2090 | * All are required to hold md->type_lock. | |
2091 | */ | |
2092 | void dm_lock_md_type(struct mapped_device *md) | |
2093 | { | |
2094 | mutex_lock(&md->type_lock); | |
2095 | } | |
2096 | ||
2097 | void dm_unlock_md_type(struct mapped_device *md) | |
2098 | { | |
2099 | mutex_unlock(&md->type_lock); | |
2100 | } | |
2101 | ||
2102 | void dm_set_md_type(struct mapped_device *md, unsigned type) | |
2103 | { | |
2104 | md->type = type; | |
2105 | } | |
2106 | ||
2107 | unsigned dm_get_md_type(struct mapped_device *md) | |
2108 | { | |
2109 | return md->type; | |
2110 | } | |
2111 | ||
4a0b4ddf MS |
2112 | /* |
2113 | * Fully initialize a request-based queue (->elevator, ->request_fn, etc). | |
2114 | */ | |
2115 | static int dm_init_request_based_queue(struct mapped_device *md) | |
2116 | { | |
2117 | struct request_queue *q = NULL; | |
2118 | ||
2119 | if (md->queue->elevator) | |
2120 | return 1; | |
2121 | ||
2122 | /* Fully initialize the queue */ | |
2123 | q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL); | |
2124 | if (!q) | |
2125 | return 0; | |
2126 | ||
2127 | md->queue = q; | |
2128 | md->saved_make_request_fn = md->queue->make_request_fn; | |
2129 | dm_init_md_queue(md); | |
2130 | blk_queue_softirq_done(md->queue, dm_softirq_done); | |
2131 | blk_queue_prep_rq(md->queue, dm_prep_fn); | |
2132 | blk_queue_lld_busy(md->queue, dm_lld_busy); | |
4a0b4ddf MS |
2133 | |
2134 | elv_register_queue(md->queue); | |
2135 | ||
2136 | return 1; | |
2137 | } | |
2138 | ||
2139 | /* | |
2140 | * Setup the DM device's queue based on md's type | |
2141 | */ | |
2142 | int dm_setup_md_queue(struct mapped_device *md) | |
2143 | { | |
2144 | if ((dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) && | |
2145 | !dm_init_request_based_queue(md)) { | |
2146 | DMWARN("Cannot initialize queue for request-based mapped device"); | |
2147 | return -EINVAL; | |
2148 | } | |
2149 | ||
2150 | return 0; | |
2151 | } | |
2152 | ||
637842cf | 2153 | static struct mapped_device *dm_find_md(dev_t dev) |
1da177e4 LT |
2154 | { |
2155 | struct mapped_device *md; | |
1da177e4 LT |
2156 | unsigned minor = MINOR(dev); |
2157 | ||
2158 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
2159 | return NULL; | |
2160 | ||
f32c10b0 | 2161 | spin_lock(&_minor_lock); |
1da177e4 LT |
2162 | |
2163 | md = idr_find(&_minor_idr, minor); | |
fba9f90e | 2164 | if (md && (md == MINOR_ALLOCED || |
f331c029 | 2165 | (MINOR(disk_devt(dm_disk(md))) != minor) || |
abdc568b | 2166 | dm_deleting_md(md) || |
17b2f66f | 2167 | test_bit(DMF_FREEING, &md->flags))) { |
637842cf | 2168 | md = NULL; |
fba9f90e JM |
2169 | goto out; |
2170 | } | |
1da177e4 | 2171 | |
fba9f90e | 2172 | out: |
f32c10b0 | 2173 | spin_unlock(&_minor_lock); |
1da177e4 | 2174 | |
637842cf DT |
2175 | return md; |
2176 | } | |
2177 | ||
d229a958 DT |
2178 | struct mapped_device *dm_get_md(dev_t dev) |
2179 | { | |
2180 | struct mapped_device *md = dm_find_md(dev); | |
2181 | ||
2182 | if (md) | |
2183 | dm_get(md); | |
2184 | ||
2185 | return md; | |
2186 | } | |
2187 | ||
9ade92a9 | 2188 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 2189 | { |
9ade92a9 | 2190 | return md->interface_ptr; |
1da177e4 LT |
2191 | } |
2192 | ||
2193 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
2194 | { | |
2195 | md->interface_ptr = ptr; | |
2196 | } | |
2197 | ||
2198 | void dm_get(struct mapped_device *md) | |
2199 | { | |
2200 | atomic_inc(&md->holders); | |
3f77316d | 2201 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1da177e4 LT |
2202 | } |
2203 | ||
72d94861 AK |
2204 | const char *dm_device_name(struct mapped_device *md) |
2205 | { | |
2206 | return md->name; | |
2207 | } | |
2208 | EXPORT_SYMBOL_GPL(dm_device_name); | |
2209 | ||
3f77316d | 2210 | static void __dm_destroy(struct mapped_device *md, bool wait) |
1da177e4 | 2211 | { |
1134e5ae | 2212 | struct dm_table *map; |
1da177e4 | 2213 | |
3f77316d | 2214 | might_sleep(); |
fba9f90e | 2215 | |
3f77316d KU |
2216 | spin_lock(&_minor_lock); |
2217 | map = dm_get_live_table(md); | |
2218 | idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); | |
2219 | set_bit(DMF_FREEING, &md->flags); | |
2220 | spin_unlock(&_minor_lock); | |
2221 | ||
2222 | if (!dm_suspended_md(md)) { | |
2223 | dm_table_presuspend_targets(map); | |
2224 | dm_table_postsuspend_targets(map); | |
1da177e4 | 2225 | } |
3f77316d KU |
2226 | |
2227 | /* | |
2228 | * Rare, but there may be I/O requests still going to complete, | |
2229 | * for example. Wait for all references to disappear. | |
2230 | * No one should increment the reference count of the mapped_device, | |
2231 | * after the mapped_device state becomes DMF_FREEING. | |
2232 | */ | |
2233 | if (wait) | |
2234 | while (atomic_read(&md->holders)) | |
2235 | msleep(1); | |
2236 | else if (atomic_read(&md->holders)) | |
2237 | DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)", | |
2238 | dm_device_name(md), atomic_read(&md->holders)); | |
2239 | ||
2240 | dm_sysfs_exit(md); | |
2241 | dm_table_put(map); | |
2242 | dm_table_destroy(__unbind(md)); | |
2243 | free_dev(md); | |
2244 | } | |
2245 | ||
2246 | void dm_destroy(struct mapped_device *md) | |
2247 | { | |
2248 | __dm_destroy(md, true); | |
2249 | } | |
2250 | ||
2251 | void dm_destroy_immediate(struct mapped_device *md) | |
2252 | { | |
2253 | __dm_destroy(md, false); | |
2254 | } | |
2255 | ||
2256 | void dm_put(struct mapped_device *md) | |
2257 | { | |
2258 | atomic_dec(&md->holders); | |
1da177e4 | 2259 | } |
79eb885c | 2260 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 2261 | |
401600df | 2262 | static int dm_wait_for_completion(struct mapped_device *md, int interruptible) |
46125c1c MB |
2263 | { |
2264 | int r = 0; | |
b44ebeb0 MP |
2265 | DECLARE_WAITQUEUE(wait, current); |
2266 | ||
2267 | dm_unplug_all(md->queue); | |
2268 | ||
2269 | add_wait_queue(&md->wait, &wait); | |
46125c1c MB |
2270 | |
2271 | while (1) { | |
401600df | 2272 | set_current_state(interruptible); |
46125c1c MB |
2273 | |
2274 | smp_mb(); | |
b4324fee | 2275 | if (!md_in_flight(md)) |
46125c1c MB |
2276 | break; |
2277 | ||
401600df MP |
2278 | if (interruptible == TASK_INTERRUPTIBLE && |
2279 | signal_pending(current)) { | |
46125c1c MB |
2280 | r = -EINTR; |
2281 | break; | |
2282 | } | |
2283 | ||
2284 | io_schedule(); | |
2285 | } | |
2286 | set_current_state(TASK_RUNNING); | |
2287 | ||
b44ebeb0 MP |
2288 | remove_wait_queue(&md->wait, &wait); |
2289 | ||
46125c1c MB |
2290 | return r; |
2291 | } | |
2292 | ||
1da177e4 LT |
2293 | /* |
2294 | * Process the deferred bios | |
2295 | */ | |
ef208587 | 2296 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 2297 | { |
ef208587 MP |
2298 | struct mapped_device *md = container_of(work, struct mapped_device, |
2299 | work); | |
6d6f10df | 2300 | struct bio *c; |
1da177e4 | 2301 | |
6a8736d1 | 2302 | down_read(&md->io_lock); |
ef208587 | 2303 | |
3b00b203 | 2304 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
2305 | spin_lock_irq(&md->deferred_lock); |
2306 | c = bio_list_pop(&md->deferred); | |
2307 | spin_unlock_irq(&md->deferred_lock); | |
2308 | ||
6a8736d1 | 2309 | if (!c) |
df12ee99 | 2310 | break; |
022c2611 | 2311 | |
6a8736d1 | 2312 | up_read(&md->io_lock); |
3b00b203 | 2313 | |
e6ee8c0b KU |
2314 | if (dm_request_based(md)) |
2315 | generic_make_request(c); | |
6a8736d1 TH |
2316 | else |
2317 | __split_and_process_bio(md, c); | |
3b00b203 | 2318 | |
6a8736d1 | 2319 | down_read(&md->io_lock); |
022c2611 | 2320 | } |
73d410c0 | 2321 | |
6a8736d1 | 2322 | up_read(&md->io_lock); |
1da177e4 LT |
2323 | } |
2324 | ||
9a1fb464 | 2325 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 2326 | { |
3b00b203 MP |
2327 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
2328 | smp_mb__after_clear_bit(); | |
53d5914f | 2329 | queue_work(md->wq, &md->work); |
304f3f6a MB |
2330 | } |
2331 | ||
1da177e4 | 2332 | /* |
042d2a9b | 2333 | * Swap in a new table, returning the old one for the caller to destroy. |
1da177e4 | 2334 | */ |
042d2a9b | 2335 | struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table) |
1da177e4 | 2336 | { |
042d2a9b | 2337 | struct dm_table *map = ERR_PTR(-EINVAL); |
754c5fc7 | 2338 | struct queue_limits limits; |
042d2a9b | 2339 | int r; |
1da177e4 | 2340 | |
e61290a4 | 2341 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
2342 | |
2343 | /* device must be suspended */ | |
4f186f8b | 2344 | if (!dm_suspended_md(md)) |
93c534ae | 2345 | goto out; |
1da177e4 | 2346 | |
754c5fc7 | 2347 | r = dm_calculate_queue_limits(table, &limits); |
042d2a9b AK |
2348 | if (r) { |
2349 | map = ERR_PTR(r); | |
754c5fc7 | 2350 | goto out; |
042d2a9b | 2351 | } |
754c5fc7 | 2352 | |
042d2a9b | 2353 | map = __bind(md, table, &limits); |
1da177e4 | 2354 | |
93c534ae | 2355 | out: |
e61290a4 | 2356 | mutex_unlock(&md->suspend_lock); |
042d2a9b | 2357 | return map; |
1da177e4 LT |
2358 | } |
2359 | ||
2360 | /* | |
2361 | * Functions to lock and unlock any filesystem running on the | |
2362 | * device. | |
2363 | */ | |
2ca3310e | 2364 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 2365 | { |
e39e2e95 | 2366 | int r; |
1da177e4 LT |
2367 | |
2368 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 2369 | |
db8fef4f | 2370 | md->frozen_sb = freeze_bdev(md->bdev); |
dfbe03f6 | 2371 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 2372 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
2373 | md->frozen_sb = NULL; |
2374 | return r; | |
dfbe03f6 AK |
2375 | } |
2376 | ||
aa8d7c2f AK |
2377 | set_bit(DMF_FROZEN, &md->flags); |
2378 | ||
1da177e4 LT |
2379 | return 0; |
2380 | } | |
2381 | ||
2ca3310e | 2382 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 2383 | { |
aa8d7c2f AK |
2384 | if (!test_bit(DMF_FROZEN, &md->flags)) |
2385 | return; | |
2386 | ||
db8fef4f | 2387 | thaw_bdev(md->bdev, md->frozen_sb); |
1da177e4 | 2388 | md->frozen_sb = NULL; |
aa8d7c2f | 2389 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
2390 | } |
2391 | ||
2392 | /* | |
2393 | * We need to be able to change a mapping table under a mounted | |
2394 | * filesystem. For example we might want to move some data in | |
2395 | * the background. Before the table can be swapped with | |
2396 | * dm_bind_table, dm_suspend must be called to flush any in | |
2397 | * flight bios and ensure that any further io gets deferred. | |
2398 | */ | |
cec47e3d KU |
2399 | /* |
2400 | * Suspend mechanism in request-based dm. | |
2401 | * | |
9f518b27 KU |
2402 | * 1. Flush all I/Os by lock_fs() if needed. |
2403 | * 2. Stop dispatching any I/O by stopping the request_queue. | |
2404 | * 3. Wait for all in-flight I/Os to be completed or requeued. | |
cec47e3d | 2405 | * |
9f518b27 | 2406 | * To abort suspend, start the request_queue. |
cec47e3d | 2407 | */ |
a3d77d35 | 2408 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) |
1da177e4 | 2409 | { |
2ca3310e | 2410 | struct dm_table *map = NULL; |
46125c1c | 2411 | int r = 0; |
a3d77d35 | 2412 | int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0; |
2e93ccc1 | 2413 | int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0; |
1da177e4 | 2414 | |
e61290a4 | 2415 | mutex_lock(&md->suspend_lock); |
2ca3310e | 2416 | |
4f186f8b | 2417 | if (dm_suspended_md(md)) { |
73d410c0 | 2418 | r = -EINVAL; |
d287483d | 2419 | goto out_unlock; |
73d410c0 | 2420 | } |
1da177e4 | 2421 | |
7c666411 | 2422 | map = dm_get_live_table(md); |
1da177e4 | 2423 | |
2e93ccc1 KU |
2424 | /* |
2425 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
2426 | * This flag is cleared before dm_suspend returns. | |
2427 | */ | |
2428 | if (noflush) | |
2429 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
2430 | ||
cf222b37 AK |
2431 | /* This does not get reverted if there's an error later. */ |
2432 | dm_table_presuspend_targets(map); | |
2433 | ||
32a926da | 2434 | /* |
9f518b27 KU |
2435 | * Flush I/O to the device. |
2436 | * Any I/O submitted after lock_fs() may not be flushed. | |
2437 | * noflush takes precedence over do_lockfs. | |
2438 | * (lock_fs() flushes I/Os and waits for them to complete.) | |
32a926da MP |
2439 | */ |
2440 | if (!noflush && do_lockfs) { | |
2441 | r = lock_fs(md); | |
2442 | if (r) | |
f431d966 | 2443 | goto out; |
aa8d7c2f | 2444 | } |
1da177e4 LT |
2445 | |
2446 | /* | |
3b00b203 MP |
2447 | * Here we must make sure that no processes are submitting requests |
2448 | * to target drivers i.e. no one may be executing | |
2449 | * __split_and_process_bio. This is called from dm_request and | |
2450 | * dm_wq_work. | |
2451 | * | |
2452 | * To get all processes out of __split_and_process_bio in dm_request, | |
2453 | * we take the write lock. To prevent any process from reentering | |
6a8736d1 TH |
2454 | * __split_and_process_bio from dm_request and quiesce the thread |
2455 | * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call | |
2456 | * flush_workqueue(md->wq). | |
1da177e4 | 2457 | */ |
2ca3310e | 2458 | down_write(&md->io_lock); |
1eb787ec | 2459 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
2ca3310e | 2460 | up_write(&md->io_lock); |
1da177e4 | 2461 | |
d0bcb878 | 2462 | /* |
29e4013d TH |
2463 | * Stop md->queue before flushing md->wq in case request-based |
2464 | * dm defers requests to md->wq from md->queue. | |
d0bcb878 | 2465 | */ |
cec47e3d | 2466 | if (dm_request_based(md)) |
9f518b27 | 2467 | stop_queue(md->queue); |
cec47e3d | 2468 | |
d0bcb878 KU |
2469 | flush_workqueue(md->wq); |
2470 | ||
1da177e4 | 2471 | /* |
3b00b203 MP |
2472 | * At this point no more requests are entering target request routines. |
2473 | * We call dm_wait_for_completion to wait for all existing requests | |
2474 | * to finish. | |
1da177e4 | 2475 | */ |
401600df | 2476 | r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE); |
1da177e4 | 2477 | |
2ca3310e | 2478 | down_write(&md->io_lock); |
6d6f10df | 2479 | if (noflush) |
022c2611 | 2480 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
94d6351e | 2481 | up_write(&md->io_lock); |
2e93ccc1 | 2482 | |
1da177e4 | 2483 | /* were we interrupted ? */ |
46125c1c | 2484 | if (r < 0) { |
9a1fb464 | 2485 | dm_queue_flush(md); |
73d410c0 | 2486 | |
cec47e3d | 2487 | if (dm_request_based(md)) |
9f518b27 | 2488 | start_queue(md->queue); |
cec47e3d | 2489 | |
2ca3310e | 2490 | unlock_fs(md); |
2e93ccc1 | 2491 | goto out; /* pushback list is already flushed, so skip flush */ |
2ca3310e | 2492 | } |
1da177e4 | 2493 | |
3b00b203 MP |
2494 | /* |
2495 | * If dm_wait_for_completion returned 0, the device is completely | |
2496 | * quiescent now. There is no request-processing activity. All new | |
2497 | * requests are being added to md->deferred list. | |
2498 | */ | |
2499 | ||
2ca3310e | 2500 | set_bit(DMF_SUSPENDED, &md->flags); |
b84b0287 | 2501 | |
4d4471cb KU |
2502 | dm_table_postsuspend_targets(map); |
2503 | ||
2ca3310e AK |
2504 | out: |
2505 | dm_table_put(map); | |
d287483d AK |
2506 | |
2507 | out_unlock: | |
e61290a4 | 2508 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 2509 | return r; |
1da177e4 LT |
2510 | } |
2511 | ||
2512 | int dm_resume(struct mapped_device *md) | |
2513 | { | |
cf222b37 | 2514 | int r = -EINVAL; |
cf222b37 | 2515 | struct dm_table *map = NULL; |
1da177e4 | 2516 | |
e61290a4 | 2517 | mutex_lock(&md->suspend_lock); |
4f186f8b | 2518 | if (!dm_suspended_md(md)) |
cf222b37 | 2519 | goto out; |
cf222b37 | 2520 | |
7c666411 | 2521 | map = dm_get_live_table(md); |
2ca3310e | 2522 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 2523 | goto out; |
1da177e4 | 2524 | |
8757b776 MB |
2525 | r = dm_table_resume_targets(map); |
2526 | if (r) | |
2527 | goto out; | |
2ca3310e | 2528 | |
9a1fb464 | 2529 | dm_queue_flush(md); |
2ca3310e | 2530 | |
cec47e3d KU |
2531 | /* |
2532 | * Flushing deferred I/Os must be done after targets are resumed | |
2533 | * so that mapping of targets can work correctly. | |
2534 | * Request-based dm is queueing the deferred I/Os in its request_queue. | |
2535 | */ | |
2536 | if (dm_request_based(md)) | |
2537 | start_queue(md->queue); | |
2538 | ||
2ca3310e AK |
2539 | unlock_fs(md); |
2540 | ||
2541 | clear_bit(DMF_SUSPENDED, &md->flags); | |
2542 | ||
1da177e4 | 2543 | dm_table_unplug_all(map); |
cf222b37 AK |
2544 | r = 0; |
2545 | out: | |
2546 | dm_table_put(map); | |
e61290a4 | 2547 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 2548 | |
cf222b37 | 2549 | return r; |
1da177e4 LT |
2550 | } |
2551 | ||
2552 | /*----------------------------------------------------------------- | |
2553 | * Event notification. | |
2554 | *---------------------------------------------------------------*/ | |
3abf85b5 | 2555 | int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action, |
60935eb2 | 2556 | unsigned cookie) |
69267a30 | 2557 | { |
60935eb2 MB |
2558 | char udev_cookie[DM_COOKIE_LENGTH]; |
2559 | char *envp[] = { udev_cookie, NULL }; | |
2560 | ||
2561 | if (!cookie) | |
3abf85b5 | 2562 | return kobject_uevent(&disk_to_dev(md->disk)->kobj, action); |
60935eb2 MB |
2563 | else { |
2564 | snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u", | |
2565 | DM_COOKIE_ENV_VAR_NAME, cookie); | |
3abf85b5 PR |
2566 | return kobject_uevent_env(&disk_to_dev(md->disk)->kobj, |
2567 | action, envp); | |
60935eb2 | 2568 | } |
69267a30 AK |
2569 | } |
2570 | ||
7a8c3d3b MA |
2571 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
2572 | { | |
2573 | return atomic_add_return(1, &md->uevent_seq); | |
2574 | } | |
2575 | ||
1da177e4 LT |
2576 | uint32_t dm_get_event_nr(struct mapped_device *md) |
2577 | { | |
2578 | return atomic_read(&md->event_nr); | |
2579 | } | |
2580 | ||
2581 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
2582 | { | |
2583 | return wait_event_interruptible(md->eventq, | |
2584 | (event_nr != atomic_read(&md->event_nr))); | |
2585 | } | |
2586 | ||
7a8c3d3b MA |
2587 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
2588 | { | |
2589 | unsigned long flags; | |
2590 | ||
2591 | spin_lock_irqsave(&md->uevent_lock, flags); | |
2592 | list_add(elist, &md->uevent_list); | |
2593 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
2594 | } | |
2595 | ||
1da177e4 LT |
2596 | /* |
2597 | * The gendisk is only valid as long as you have a reference | |
2598 | * count on 'md'. | |
2599 | */ | |
2600 | struct gendisk *dm_disk(struct mapped_device *md) | |
2601 | { | |
2602 | return md->disk; | |
2603 | } | |
2604 | ||
784aae73 MB |
2605 | struct kobject *dm_kobject(struct mapped_device *md) |
2606 | { | |
2607 | return &md->kobj; | |
2608 | } | |
2609 | ||
2610 | /* | |
2611 | * struct mapped_device should not be exported outside of dm.c | |
2612 | * so use this check to verify that kobj is part of md structure | |
2613 | */ | |
2614 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) | |
2615 | { | |
2616 | struct mapped_device *md; | |
2617 | ||
2618 | md = container_of(kobj, struct mapped_device, kobj); | |
2619 | if (&md->kobj != kobj) | |
2620 | return NULL; | |
2621 | ||
4d89b7b4 | 2622 | if (test_bit(DMF_FREEING, &md->flags) || |
432a212c | 2623 | dm_deleting_md(md)) |
4d89b7b4 MB |
2624 | return NULL; |
2625 | ||
784aae73 MB |
2626 | dm_get(md); |
2627 | return md; | |
2628 | } | |
2629 | ||
4f186f8b | 2630 | int dm_suspended_md(struct mapped_device *md) |
1da177e4 LT |
2631 | { |
2632 | return test_bit(DMF_SUSPENDED, &md->flags); | |
2633 | } | |
2634 | ||
64dbce58 KU |
2635 | int dm_suspended(struct dm_target *ti) |
2636 | { | |
ecdb2e25 | 2637 | return dm_suspended_md(dm_table_get_md(ti->table)); |
64dbce58 KU |
2638 | } |
2639 | EXPORT_SYMBOL_GPL(dm_suspended); | |
2640 | ||
2e93ccc1 KU |
2641 | int dm_noflush_suspending(struct dm_target *ti) |
2642 | { | |
ecdb2e25 | 2643 | return __noflush_suspending(dm_table_get_md(ti->table)); |
2e93ccc1 KU |
2644 | } |
2645 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
2646 | ||
e6ee8c0b KU |
2647 | struct dm_md_mempools *dm_alloc_md_mempools(unsigned type) |
2648 | { | |
2649 | struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL); | |
2650 | ||
2651 | if (!pools) | |
2652 | return NULL; | |
2653 | ||
2654 | pools->io_pool = (type == DM_TYPE_BIO_BASED) ? | |
2655 | mempool_create_slab_pool(MIN_IOS, _io_cache) : | |
2656 | mempool_create_slab_pool(MIN_IOS, _rq_bio_info_cache); | |
2657 | if (!pools->io_pool) | |
2658 | goto free_pools_and_out; | |
2659 | ||
2660 | pools->tio_pool = (type == DM_TYPE_BIO_BASED) ? | |
2661 | mempool_create_slab_pool(MIN_IOS, _tio_cache) : | |
2662 | mempool_create_slab_pool(MIN_IOS, _rq_tio_cache); | |
2663 | if (!pools->tio_pool) | |
2664 | goto free_io_pool_and_out; | |
2665 | ||
2666 | pools->bs = (type == DM_TYPE_BIO_BASED) ? | |
2667 | bioset_create(16, 0) : bioset_create(MIN_IOS, 0); | |
2668 | if (!pools->bs) | |
2669 | goto free_tio_pool_and_out; | |
2670 | ||
2671 | return pools; | |
2672 | ||
2673 | free_tio_pool_and_out: | |
2674 | mempool_destroy(pools->tio_pool); | |
2675 | ||
2676 | free_io_pool_and_out: | |
2677 | mempool_destroy(pools->io_pool); | |
2678 | ||
2679 | free_pools_and_out: | |
2680 | kfree(pools); | |
2681 | ||
2682 | return NULL; | |
2683 | } | |
2684 | ||
2685 | void dm_free_md_mempools(struct dm_md_mempools *pools) | |
2686 | { | |
2687 | if (!pools) | |
2688 | return; | |
2689 | ||
2690 | if (pools->io_pool) | |
2691 | mempool_destroy(pools->io_pool); | |
2692 | ||
2693 | if (pools->tio_pool) | |
2694 | mempool_destroy(pools->tio_pool); | |
2695 | ||
2696 | if (pools->bs) | |
2697 | bioset_free(pools->bs); | |
2698 | ||
2699 | kfree(pools); | |
2700 | } | |
2701 | ||
83d5cde4 | 2702 | static const struct block_device_operations dm_blk_dops = { |
1da177e4 LT |
2703 | .open = dm_blk_open, |
2704 | .release = dm_blk_close, | |
aa129a22 | 2705 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 2706 | .getgeo = dm_blk_getgeo, |
1da177e4 LT |
2707 | .owner = THIS_MODULE |
2708 | }; | |
2709 | ||
2710 | EXPORT_SYMBOL(dm_get_mapinfo); | |
2711 | ||
2712 | /* | |
2713 | * module hooks | |
2714 | */ | |
2715 | module_init(dm_init); | |
2716 | module_exit(dm_exit); | |
2717 | ||
2718 | module_param(major, uint, 0); | |
2719 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
2720 | MODULE_DESCRIPTION(DM_NAME " driver"); | |
2721 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
2722 | MODULE_LICENSE("GPL"); |