]>
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 | ||
4cc96131 MS |
8 | #include "dm-core.h" |
9 | #include "dm-rq.h" | |
51e5b2bd | 10 | #include "dm-uevent.h" |
1da177e4 LT |
11 | |
12 | #include <linux/init.h> | |
13 | #include <linux/module.h> | |
48c9c27b | 14 | #include <linux/mutex.h> |
174cd4b1 | 15 | #include <linux/sched/signal.h> |
1da177e4 LT |
16 | #include <linux/blkpg.h> |
17 | #include <linux/bio.h> | |
1da177e4 | 18 | #include <linux/mempool.h> |
f26c5719 | 19 | #include <linux/dax.h> |
1da177e4 LT |
20 | #include <linux/slab.h> |
21 | #include <linux/idr.h> | |
7e026c8c | 22 | #include <linux/uio.h> |
3ac51e74 | 23 | #include <linux/hdreg.h> |
3f77316d | 24 | #include <linux/delay.h> |
ffcc3936 | 25 | #include <linux/wait.h> |
71cdb697 | 26 | #include <linux/pr.h> |
b0b4d7c6 | 27 | #include <linux/refcount.h> |
55782138 | 28 | |
72d94861 AK |
29 | #define DM_MSG_PREFIX "core" |
30 | ||
60935eb2 MB |
31 | /* |
32 | * Cookies are numeric values sent with CHANGE and REMOVE | |
33 | * uevents while resuming, removing or renaming the device. | |
34 | */ | |
35 | #define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE" | |
36 | #define DM_COOKIE_LENGTH 24 | |
37 | ||
1da177e4 LT |
38 | static const char *_name = DM_NAME; |
39 | ||
40 | static unsigned int major = 0; | |
41 | static unsigned int _major = 0; | |
42 | ||
d15b774c AK |
43 | static DEFINE_IDR(_minor_idr); |
44 | ||
f32c10b0 | 45 | static DEFINE_SPINLOCK(_minor_lock); |
2c140a24 MP |
46 | |
47 | static void do_deferred_remove(struct work_struct *w); | |
48 | ||
49 | static DECLARE_WORK(deferred_remove_work, do_deferred_remove); | |
50 | ||
acfe0ad7 MP |
51 | static struct workqueue_struct *deferred_remove_workqueue; |
52 | ||
93e6442c MP |
53 | atomic_t dm_global_event_nr = ATOMIC_INIT(0); |
54 | DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); | |
55 | ||
62e08243 MP |
56 | void dm_issue_global_event(void) |
57 | { | |
58 | atomic_inc(&dm_global_event_nr); | |
59 | wake_up(&dm_global_eventq); | |
60 | } | |
61 | ||
64f52b0e MS |
62 | /* |
63 | * One of these is allocated (on-stack) per original bio. | |
64 | */ | |
65 | struct clone_info { | |
66 | struct mapped_device *md; | |
67 | struct dm_table *map; | |
68 | struct bio *bio; | |
69 | struct dm_io *io; | |
70 | sector_t sector; | |
71 | unsigned sector_count; | |
72 | }; | |
73 | ||
74 | /* | |
75 | * One of these is allocated per clone bio. | |
76 | */ | |
77 | #define DM_TIO_MAGIC 7282014 | |
78 | struct dm_target_io { | |
79 | unsigned magic; | |
80 | struct dm_io *io; | |
81 | struct dm_target *ti; | |
82 | unsigned target_bio_nr; | |
83 | unsigned *len_ptr; | |
84 | bool inside_dm_io; | |
85 | struct bio clone; | |
86 | }; | |
87 | ||
1da177e4 | 88 | /* |
745dc570 | 89 | * One of these is allocated per original bio. |
64f52b0e | 90 | * It contains the first clone used for that original. |
1da177e4 | 91 | */ |
64f52b0e | 92 | #define DM_IO_MAGIC 5191977 |
1da177e4 | 93 | struct dm_io { |
64f52b0e | 94 | unsigned magic; |
1da177e4 | 95 | struct mapped_device *md; |
4e4cbee9 | 96 | blk_status_t status; |
1da177e4 | 97 | atomic_t io_count; |
745dc570 | 98 | struct bio *orig_bio; |
3eaf840e | 99 | unsigned long start_time; |
f88fb981 | 100 | spinlock_t endio_lock; |
fd2ed4d2 | 101 | struct dm_stats_aux stats_aux; |
64f52b0e MS |
102 | /* last member of dm_target_io is 'struct bio' */ |
103 | struct dm_target_io tio; | |
1da177e4 LT |
104 | }; |
105 | ||
64f52b0e MS |
106 | void *dm_per_bio_data(struct bio *bio, size_t data_size) |
107 | { | |
108 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); | |
109 | if (!tio->inside_dm_io) | |
110 | return (char *)bio - offsetof(struct dm_target_io, clone) - data_size; | |
111 | return (char *)bio - offsetof(struct dm_target_io, clone) - offsetof(struct dm_io, tio) - data_size; | |
112 | } | |
113 | EXPORT_SYMBOL_GPL(dm_per_bio_data); | |
114 | ||
115 | struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size) | |
116 | { | |
117 | struct dm_io *io = (struct dm_io *)((char *)data + data_size); | |
118 | if (io->magic == DM_IO_MAGIC) | |
119 | return (struct bio *)((char *)io + offsetof(struct dm_io, tio) + offsetof(struct dm_target_io, clone)); | |
120 | BUG_ON(io->magic != DM_TIO_MAGIC); | |
121 | return (struct bio *)((char *)io + offsetof(struct dm_target_io, clone)); | |
122 | } | |
123 | EXPORT_SYMBOL_GPL(dm_bio_from_per_bio_data); | |
124 | ||
125 | unsigned dm_bio_get_target_bio_nr(const struct bio *bio) | |
126 | { | |
127 | return container_of(bio, struct dm_target_io, clone)->target_bio_nr; | |
128 | } | |
129 | EXPORT_SYMBOL_GPL(dm_bio_get_target_bio_nr); | |
130 | ||
ba61fdd1 JM |
131 | #define MINOR_ALLOCED ((void *)-1) |
132 | ||
1da177e4 LT |
133 | /* |
134 | * Bits for the md->flags field. | |
135 | */ | |
1eb787ec | 136 | #define DMF_BLOCK_IO_FOR_SUSPEND 0 |
1da177e4 | 137 | #define DMF_SUSPENDED 1 |
aa8d7c2f | 138 | #define DMF_FROZEN 2 |
fba9f90e | 139 | #define DMF_FREEING 3 |
5c6bd75d | 140 | #define DMF_DELETING 4 |
2e93ccc1 | 141 | #define DMF_NOFLUSH_SUSPENDING 5 |
8ae12666 KO |
142 | #define DMF_DEFERRED_REMOVE 6 |
143 | #define DMF_SUSPENDED_INTERNALLY 7 | |
1da177e4 | 144 | |
115485e8 | 145 | #define DM_NUMA_NODE NUMA_NO_NODE |
115485e8 | 146 | static int dm_numa_node = DM_NUMA_NODE; |
faad87df | 147 | |
e6ee8c0b KU |
148 | /* |
149 | * For mempools pre-allocation at the table loading time. | |
150 | */ | |
151 | struct dm_md_mempools { | |
e6ee8c0b | 152 | struct bio_set *bs; |
64f52b0e | 153 | struct bio_set *io_bs; |
e6ee8c0b KU |
154 | }; |
155 | ||
86f1152b BM |
156 | struct table_device { |
157 | struct list_head list; | |
b0b4d7c6 | 158 | refcount_t count; |
86f1152b BM |
159 | struct dm_dev dm_dev; |
160 | }; | |
161 | ||
8fbf26ad | 162 | static struct kmem_cache *_rq_tio_cache; |
1ae49ea2 | 163 | static struct kmem_cache *_rq_cache; |
94818742 | 164 | |
e8603136 MS |
165 | /* |
166 | * Bio-based DM's mempools' reserved IOs set by the user. | |
167 | */ | |
4cc96131 | 168 | #define RESERVED_BIO_BASED_IOS 16 |
e8603136 MS |
169 | static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS; |
170 | ||
115485e8 MS |
171 | static int __dm_get_module_param_int(int *module_param, int min, int max) |
172 | { | |
6aa7de05 | 173 | int param = READ_ONCE(*module_param); |
115485e8 MS |
174 | int modified_param = 0; |
175 | bool modified = true; | |
176 | ||
177 | if (param < min) | |
178 | modified_param = min; | |
179 | else if (param > max) | |
180 | modified_param = max; | |
181 | else | |
182 | modified = false; | |
183 | ||
184 | if (modified) { | |
185 | (void)cmpxchg(module_param, param, modified_param); | |
186 | param = modified_param; | |
187 | } | |
188 | ||
189 | return param; | |
190 | } | |
191 | ||
4cc96131 MS |
192 | unsigned __dm_get_module_param(unsigned *module_param, |
193 | unsigned def, unsigned max) | |
f4790826 | 194 | { |
6aa7de05 | 195 | unsigned param = READ_ONCE(*module_param); |
09c2d531 | 196 | unsigned modified_param = 0; |
f4790826 | 197 | |
09c2d531 MS |
198 | if (!param) |
199 | modified_param = def; | |
200 | else if (param > max) | |
201 | modified_param = max; | |
f4790826 | 202 | |
09c2d531 MS |
203 | if (modified_param) { |
204 | (void)cmpxchg(module_param, param, modified_param); | |
205 | param = modified_param; | |
f4790826 MS |
206 | } |
207 | ||
09c2d531 | 208 | return param; |
f4790826 MS |
209 | } |
210 | ||
e8603136 MS |
211 | unsigned dm_get_reserved_bio_based_ios(void) |
212 | { | |
09c2d531 | 213 | return __dm_get_module_param(&reserved_bio_based_ios, |
4cc96131 | 214 | RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS); |
e8603136 MS |
215 | } |
216 | EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios); | |
217 | ||
115485e8 MS |
218 | static unsigned dm_get_numa_node(void) |
219 | { | |
220 | return __dm_get_module_param_int(&dm_numa_node, | |
221 | DM_NUMA_NODE, num_online_nodes() - 1); | |
222 | } | |
223 | ||
1da177e4 LT |
224 | static int __init local_init(void) |
225 | { | |
51157b4a | 226 | int r = -ENOMEM; |
1da177e4 | 227 | |
8fbf26ad KU |
228 | _rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0); |
229 | if (!_rq_tio_cache) | |
dde1e1ec | 230 | return r; |
8fbf26ad | 231 | |
eca7ee6d | 232 | _rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request), |
1ae49ea2 MS |
233 | __alignof__(struct request), 0, NULL); |
234 | if (!_rq_cache) | |
235 | goto out_free_rq_tio_cache; | |
236 | ||
51e5b2bd | 237 | r = dm_uevent_init(); |
51157b4a | 238 | if (r) |
1ae49ea2 | 239 | goto out_free_rq_cache; |
51e5b2bd | 240 | |
acfe0ad7 MP |
241 | deferred_remove_workqueue = alloc_workqueue("kdmremove", WQ_UNBOUND, 1); |
242 | if (!deferred_remove_workqueue) { | |
243 | r = -ENOMEM; | |
244 | goto out_uevent_exit; | |
245 | } | |
246 | ||
1da177e4 LT |
247 | _major = major; |
248 | r = register_blkdev(_major, _name); | |
51157b4a | 249 | if (r < 0) |
acfe0ad7 | 250 | goto out_free_workqueue; |
1da177e4 LT |
251 | |
252 | if (!_major) | |
253 | _major = r; | |
254 | ||
255 | return 0; | |
51157b4a | 256 | |
acfe0ad7 MP |
257 | out_free_workqueue: |
258 | destroy_workqueue(deferred_remove_workqueue); | |
51157b4a KU |
259 | out_uevent_exit: |
260 | dm_uevent_exit(); | |
1ae49ea2 MS |
261 | out_free_rq_cache: |
262 | kmem_cache_destroy(_rq_cache); | |
8fbf26ad KU |
263 | out_free_rq_tio_cache: |
264 | kmem_cache_destroy(_rq_tio_cache); | |
51157b4a KU |
265 | |
266 | return r; | |
1da177e4 LT |
267 | } |
268 | ||
269 | static void local_exit(void) | |
270 | { | |
2c140a24 | 271 | flush_scheduled_work(); |
acfe0ad7 | 272 | destroy_workqueue(deferred_remove_workqueue); |
2c140a24 | 273 | |
1ae49ea2 | 274 | kmem_cache_destroy(_rq_cache); |
8fbf26ad | 275 | kmem_cache_destroy(_rq_tio_cache); |
00d59405 | 276 | unregister_blkdev(_major, _name); |
51e5b2bd | 277 | dm_uevent_exit(); |
1da177e4 LT |
278 | |
279 | _major = 0; | |
280 | ||
281 | DMINFO("cleaned up"); | |
282 | } | |
283 | ||
b9249e55 | 284 | static int (*_inits[])(void) __initdata = { |
1da177e4 LT |
285 | local_init, |
286 | dm_target_init, | |
287 | dm_linear_init, | |
288 | dm_stripe_init, | |
952b3557 | 289 | dm_io_init, |
945fa4d2 | 290 | dm_kcopyd_init, |
1da177e4 | 291 | dm_interface_init, |
fd2ed4d2 | 292 | dm_statistics_init, |
1da177e4 LT |
293 | }; |
294 | ||
b9249e55 | 295 | static void (*_exits[])(void) = { |
1da177e4 LT |
296 | local_exit, |
297 | dm_target_exit, | |
298 | dm_linear_exit, | |
299 | dm_stripe_exit, | |
952b3557 | 300 | dm_io_exit, |
945fa4d2 | 301 | dm_kcopyd_exit, |
1da177e4 | 302 | dm_interface_exit, |
fd2ed4d2 | 303 | dm_statistics_exit, |
1da177e4 LT |
304 | }; |
305 | ||
306 | static int __init dm_init(void) | |
307 | { | |
308 | const int count = ARRAY_SIZE(_inits); | |
309 | ||
310 | int r, i; | |
311 | ||
312 | for (i = 0; i < count; i++) { | |
313 | r = _inits[i](); | |
314 | if (r) | |
315 | goto bad; | |
316 | } | |
317 | ||
318 | return 0; | |
319 | ||
320 | bad: | |
321 | while (i--) | |
322 | _exits[i](); | |
323 | ||
324 | return r; | |
325 | } | |
326 | ||
327 | static void __exit dm_exit(void) | |
328 | { | |
329 | int i = ARRAY_SIZE(_exits); | |
330 | ||
331 | while (i--) | |
332 | _exits[i](); | |
d15b774c AK |
333 | |
334 | /* | |
335 | * Should be empty by this point. | |
336 | */ | |
d15b774c | 337 | idr_destroy(&_minor_idr); |
1da177e4 LT |
338 | } |
339 | ||
340 | /* | |
341 | * Block device functions | |
342 | */ | |
432a212c MA |
343 | int dm_deleting_md(struct mapped_device *md) |
344 | { | |
345 | return test_bit(DMF_DELETING, &md->flags); | |
346 | } | |
347 | ||
fe5f9f2c | 348 | static int dm_blk_open(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
349 | { |
350 | struct mapped_device *md; | |
351 | ||
fba9f90e JM |
352 | spin_lock(&_minor_lock); |
353 | ||
fe5f9f2c | 354 | md = bdev->bd_disk->private_data; |
fba9f90e JM |
355 | if (!md) |
356 | goto out; | |
357 | ||
5c6bd75d | 358 | if (test_bit(DMF_FREEING, &md->flags) || |
432a212c | 359 | dm_deleting_md(md)) { |
fba9f90e JM |
360 | md = NULL; |
361 | goto out; | |
362 | } | |
363 | ||
1da177e4 | 364 | dm_get(md); |
5c6bd75d | 365 | atomic_inc(&md->open_count); |
fba9f90e JM |
366 | out: |
367 | spin_unlock(&_minor_lock); | |
368 | ||
369 | return md ? 0 : -ENXIO; | |
1da177e4 LT |
370 | } |
371 | ||
db2a144b | 372 | static void dm_blk_close(struct gendisk *disk, fmode_t mode) |
1da177e4 | 373 | { |
63a4f065 | 374 | struct mapped_device *md; |
6e9624b8 | 375 | |
4a1aeb98 MB |
376 | spin_lock(&_minor_lock); |
377 | ||
63a4f065 MS |
378 | md = disk->private_data; |
379 | if (WARN_ON(!md)) | |
380 | goto out; | |
381 | ||
2c140a24 MP |
382 | if (atomic_dec_and_test(&md->open_count) && |
383 | (test_bit(DMF_DEFERRED_REMOVE, &md->flags))) | |
acfe0ad7 | 384 | queue_work(deferred_remove_workqueue, &deferred_remove_work); |
2c140a24 | 385 | |
1da177e4 | 386 | dm_put(md); |
63a4f065 | 387 | out: |
4a1aeb98 | 388 | spin_unlock(&_minor_lock); |
1da177e4 LT |
389 | } |
390 | ||
5c6bd75d AK |
391 | int dm_open_count(struct mapped_device *md) |
392 | { | |
393 | return atomic_read(&md->open_count); | |
394 | } | |
395 | ||
396 | /* | |
397 | * Guarantees nothing is using the device before it's deleted. | |
398 | */ | |
2c140a24 | 399 | int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred) |
5c6bd75d AK |
400 | { |
401 | int r = 0; | |
402 | ||
403 | spin_lock(&_minor_lock); | |
404 | ||
2c140a24 | 405 | if (dm_open_count(md)) { |
5c6bd75d | 406 | r = -EBUSY; |
2c140a24 MP |
407 | if (mark_deferred) |
408 | set_bit(DMF_DEFERRED_REMOVE, &md->flags); | |
409 | } else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags)) | |
410 | r = -EEXIST; | |
5c6bd75d AK |
411 | else |
412 | set_bit(DMF_DELETING, &md->flags); | |
413 | ||
414 | spin_unlock(&_minor_lock); | |
415 | ||
416 | return r; | |
417 | } | |
418 | ||
2c140a24 MP |
419 | int dm_cancel_deferred_remove(struct mapped_device *md) |
420 | { | |
421 | int r = 0; | |
422 | ||
423 | spin_lock(&_minor_lock); | |
424 | ||
425 | if (test_bit(DMF_DELETING, &md->flags)) | |
426 | r = -EBUSY; | |
427 | else | |
428 | clear_bit(DMF_DEFERRED_REMOVE, &md->flags); | |
429 | ||
430 | spin_unlock(&_minor_lock); | |
431 | ||
432 | return r; | |
433 | } | |
434 | ||
435 | static void do_deferred_remove(struct work_struct *w) | |
436 | { | |
437 | dm_deferred_remove(); | |
438 | } | |
439 | ||
fd2ed4d2 MP |
440 | sector_t dm_get_size(struct mapped_device *md) |
441 | { | |
442 | return get_capacity(md->disk); | |
443 | } | |
444 | ||
9974fa2c MS |
445 | struct request_queue *dm_get_md_queue(struct mapped_device *md) |
446 | { | |
447 | return md->queue; | |
448 | } | |
449 | ||
fd2ed4d2 MP |
450 | struct dm_stats *dm_get_stats(struct mapped_device *md) |
451 | { | |
452 | return &md->stats; | |
453 | } | |
454 | ||
3ac51e74 DW |
455 | static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
456 | { | |
457 | struct mapped_device *md = bdev->bd_disk->private_data; | |
458 | ||
459 | return dm_get_geometry(md, geo); | |
460 | } | |
461 | ||
956a4025 MS |
462 | static int dm_grab_bdev_for_ioctl(struct mapped_device *md, |
463 | struct block_device **bdev, | |
464 | fmode_t *mode) | |
aa129a22 | 465 | { |
66482026 | 466 | struct dm_target *tgt; |
6c182cd8 | 467 | struct dm_table *map; |
956a4025 | 468 | int srcu_idx, r; |
aa129a22 | 469 | |
6c182cd8 | 470 | retry: |
e56f81e0 | 471 | r = -ENOTTY; |
956a4025 | 472 | map = dm_get_live_table(md, &srcu_idx); |
aa129a22 MB |
473 | if (!map || !dm_table_get_size(map)) |
474 | goto out; | |
475 | ||
476 | /* We only support devices that have a single target */ | |
477 | if (dm_table_get_num_targets(map) != 1) | |
478 | goto out; | |
479 | ||
66482026 MS |
480 | tgt = dm_table_get_target(map, 0); |
481 | if (!tgt->type->prepare_ioctl) | |
4d341d82 | 482 | goto out; |
aa129a22 | 483 | |
4f186f8b | 484 | if (dm_suspended_md(md)) { |
aa129a22 MB |
485 | r = -EAGAIN; |
486 | goto out; | |
487 | } | |
488 | ||
66482026 | 489 | r = tgt->type->prepare_ioctl(tgt, bdev, mode); |
e56f81e0 CH |
490 | if (r < 0) |
491 | goto out; | |
aa129a22 | 492 | |
956a4025 MS |
493 | bdgrab(*bdev); |
494 | dm_put_live_table(md, srcu_idx); | |
e56f81e0 | 495 | return r; |
aa129a22 | 496 | |
aa129a22 | 497 | out: |
956a4025 | 498 | dm_put_live_table(md, srcu_idx); |
5bbbfdf6 | 499 | if (r == -ENOTCONN && !fatal_signal_pending(current)) { |
6c182cd8 HR |
500 | msleep(10); |
501 | goto retry; | |
502 | } | |
e56f81e0 CH |
503 | return r; |
504 | } | |
505 | ||
506 | static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode, | |
507 | unsigned int cmd, unsigned long arg) | |
508 | { | |
509 | struct mapped_device *md = bdev->bd_disk->private_data; | |
956a4025 | 510 | int r; |
e56f81e0 | 511 | |
956a4025 | 512 | r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); |
e56f81e0 CH |
513 | if (r < 0) |
514 | return r; | |
6c182cd8 | 515 | |
e56f81e0 CH |
516 | if (r > 0) { |
517 | /* | |
e980f623 CH |
518 | * Target determined this ioctl is being issued against a |
519 | * subset of the parent bdev; require extra privileges. | |
e56f81e0 | 520 | */ |
e980f623 CH |
521 | if (!capable(CAP_SYS_RAWIO)) { |
522 | DMWARN_LIMIT( | |
523 | "%s: sending ioctl %x to DM device without required privilege.", | |
524 | current->comm, cmd); | |
525 | r = -ENOIOCTLCMD; | |
e56f81e0 | 526 | goto out; |
e980f623 | 527 | } |
e56f81e0 | 528 | } |
6c182cd8 | 529 | |
66482026 | 530 | r = __blkdev_driver_ioctl(bdev, mode, cmd, arg); |
e56f81e0 | 531 | out: |
956a4025 | 532 | bdput(bdev); |
aa129a22 MB |
533 | return r; |
534 | } | |
535 | ||
028867ac | 536 | static struct dm_io *alloc_io(struct mapped_device *md) |
1da177e4 | 537 | { |
64f52b0e MS |
538 | struct dm_io *io; |
539 | struct dm_target_io *tio; | |
540 | struct bio *clone; | |
541 | ||
542 | clone = bio_alloc_bioset(GFP_NOIO, 0, md->io_bs); | |
543 | if (!clone) | |
544 | return NULL; | |
545 | ||
546 | tio = container_of(clone, struct dm_target_io, clone); | |
547 | tio->inside_dm_io = true; | |
548 | tio->io = NULL; | |
549 | ||
550 | io = container_of(tio, struct dm_io, tio); | |
551 | io->magic = DM_IO_MAGIC; | |
552 | ||
553 | return io; | |
1da177e4 LT |
554 | } |
555 | ||
028867ac | 556 | static void free_io(struct mapped_device *md, struct dm_io *io) |
1da177e4 | 557 | { |
64f52b0e MS |
558 | bio_put(&io->tio.clone); |
559 | } | |
560 | ||
561 | static struct dm_target_io *alloc_tio(struct clone_info *ci, struct dm_target *ti, | |
562 | unsigned target_bio_nr, gfp_t gfp_mask) | |
563 | { | |
564 | struct dm_target_io *tio; | |
565 | ||
566 | if (!ci->io->tio.io) { | |
567 | /* the dm_target_io embedded in ci->io is available */ | |
568 | tio = &ci->io->tio; | |
569 | } else { | |
570 | struct bio *clone = bio_alloc_bioset(gfp_mask, 0, ci->md->bs); | |
571 | if (!clone) | |
572 | return NULL; | |
573 | ||
574 | tio = container_of(clone, struct dm_target_io, clone); | |
575 | tio->inside_dm_io = false; | |
576 | } | |
577 | ||
578 | tio->magic = DM_TIO_MAGIC; | |
579 | tio->io = ci->io; | |
580 | tio->ti = ti; | |
581 | tio->target_bio_nr = target_bio_nr; | |
582 | ||
583 | return tio; | |
1da177e4 LT |
584 | } |
585 | ||
cfae7529 | 586 | static void free_tio(struct dm_target_io *tio) |
1da177e4 | 587 | { |
64f52b0e MS |
588 | if (tio->inside_dm_io) |
589 | return; | |
dba14160 | 590 | bio_put(&tio->clone); |
1da177e4 LT |
591 | } |
592 | ||
4cc96131 | 593 | int md_in_flight(struct mapped_device *md) |
90abb8c4 KU |
594 | { |
595 | return atomic_read(&md->pending[READ]) + | |
596 | atomic_read(&md->pending[WRITE]); | |
597 | } | |
598 | ||
3eaf840e JNN |
599 | static void start_io_acct(struct dm_io *io) |
600 | { | |
601 | struct mapped_device *md = io->md; | |
745dc570 | 602 | struct bio *bio = io->orig_bio; |
c9959059 | 603 | int cpu; |
fd2ed4d2 | 604 | int rw = bio_data_dir(bio); |
3eaf840e JNN |
605 | |
606 | io->start_time = jiffies; | |
607 | ||
074a7aca | 608 | cpu = part_stat_lock(); |
d62e26b3 | 609 | part_round_stats(md->queue, cpu, &dm_disk(md)->part0); |
074a7aca | 610 | part_stat_unlock(); |
1e9bb880 SL |
611 | atomic_set(&dm_disk(md)->part0.in_flight[rw], |
612 | atomic_inc_return(&md->pending[rw])); | |
fd2ed4d2 MP |
613 | |
614 | if (unlikely(dm_stats_used(&md->stats))) | |
528ec5ab MC |
615 | dm_stats_account_io(&md->stats, bio_data_dir(bio), |
616 | bio->bi_iter.bi_sector, bio_sectors(bio), | |
617 | false, 0, &io->stats_aux); | |
3eaf840e JNN |
618 | } |
619 | ||
d221d2e7 | 620 | static void end_io_acct(struct dm_io *io) |
3eaf840e JNN |
621 | { |
622 | struct mapped_device *md = io->md; | |
745dc570 | 623 | struct bio *bio = io->orig_bio; |
3eaf840e | 624 | unsigned long duration = jiffies - io->start_time; |
18c0b223 | 625 | int pending; |
3eaf840e JNN |
626 | int rw = bio_data_dir(bio); |
627 | ||
d62e26b3 | 628 | generic_end_io_acct(md->queue, rw, &dm_disk(md)->part0, io->start_time); |
3eaf840e | 629 | |
fd2ed4d2 | 630 | if (unlikely(dm_stats_used(&md->stats))) |
528ec5ab MC |
631 | dm_stats_account_io(&md->stats, bio_data_dir(bio), |
632 | bio->bi_iter.bi_sector, bio_sectors(bio), | |
633 | true, duration, &io->stats_aux); | |
fd2ed4d2 | 634 | |
af7e466a MP |
635 | /* |
636 | * After this is decremented the bio must not be touched if it is | |
d87f4c14 | 637 | * a flush. |
af7e466a | 638 | */ |
1e9bb880 SL |
639 | pending = atomic_dec_return(&md->pending[rw]); |
640 | atomic_set(&dm_disk(md)->part0.in_flight[rw], pending); | |
316d315b | 641 | pending += atomic_read(&md->pending[rw^0x1]); |
3eaf840e | 642 | |
d221d2e7 MP |
643 | /* nudge anyone waiting on suspend queue */ |
644 | if (!pending) | |
645 | wake_up(&md->wait); | |
3eaf840e JNN |
646 | } |
647 | ||
1da177e4 LT |
648 | /* |
649 | * Add the bio to the list of deferred io. | |
650 | */ | |
92c63902 | 651 | static void queue_io(struct mapped_device *md, struct bio *bio) |
1da177e4 | 652 | { |
05447420 | 653 | unsigned long flags; |
1da177e4 | 654 | |
05447420 | 655 | spin_lock_irqsave(&md->deferred_lock, flags); |
1da177e4 | 656 | bio_list_add(&md->deferred, bio); |
05447420 | 657 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
6a8736d1 | 658 | queue_work(md->wq, &md->work); |
1da177e4 LT |
659 | } |
660 | ||
661 | /* | |
662 | * Everyone (including functions in this file), should use this | |
663 | * function to access the md->map field, and make sure they call | |
83d5e5b0 | 664 | * dm_put_live_table() when finished. |
1da177e4 | 665 | */ |
83d5e5b0 | 666 | struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier) |
1da177e4 | 667 | { |
83d5e5b0 MP |
668 | *srcu_idx = srcu_read_lock(&md->io_barrier); |
669 | ||
670 | return srcu_dereference(md->map, &md->io_barrier); | |
671 | } | |
1da177e4 | 672 | |
83d5e5b0 MP |
673 | void dm_put_live_table(struct mapped_device *md, int srcu_idx) __releases(md->io_barrier) |
674 | { | |
675 | srcu_read_unlock(&md->io_barrier, srcu_idx); | |
676 | } | |
677 | ||
678 | void dm_sync_table(struct mapped_device *md) | |
679 | { | |
680 | synchronize_srcu(&md->io_barrier); | |
681 | synchronize_rcu_expedited(); | |
682 | } | |
683 | ||
684 | /* | |
685 | * A fast alternative to dm_get_live_table/dm_put_live_table. | |
686 | * The caller must not block between these two functions. | |
687 | */ | |
688 | static struct dm_table *dm_get_live_table_fast(struct mapped_device *md) __acquires(RCU) | |
689 | { | |
690 | rcu_read_lock(); | |
691 | return rcu_dereference(md->map); | |
692 | } | |
1da177e4 | 693 | |
83d5e5b0 MP |
694 | static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) |
695 | { | |
696 | rcu_read_unlock(); | |
1da177e4 LT |
697 | } |
698 | ||
86f1152b BM |
699 | /* |
700 | * Open a table device so we can use it as a map destination. | |
701 | */ | |
702 | static int open_table_device(struct table_device *td, dev_t dev, | |
703 | struct mapped_device *md) | |
704 | { | |
705 | static char *_claim_ptr = "I belong to device-mapper"; | |
706 | struct block_device *bdev; | |
707 | ||
708 | int r; | |
709 | ||
710 | BUG_ON(td->dm_dev.bdev); | |
711 | ||
712 | bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _claim_ptr); | |
713 | if (IS_ERR(bdev)) | |
714 | return PTR_ERR(bdev); | |
715 | ||
716 | r = bd_link_disk_holder(bdev, dm_disk(md)); | |
717 | if (r) { | |
718 | blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); | |
719 | return r; | |
720 | } | |
721 | ||
722 | td->dm_dev.bdev = bdev; | |
817bf402 | 723 | td->dm_dev.dax_dev = dax_get_by_host(bdev->bd_disk->disk_name); |
86f1152b BM |
724 | return 0; |
725 | } | |
726 | ||
727 | /* | |
728 | * Close a table device that we've been using. | |
729 | */ | |
730 | static void close_table_device(struct table_device *td, struct mapped_device *md) | |
731 | { | |
732 | if (!td->dm_dev.bdev) | |
733 | return; | |
734 | ||
735 | bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); | |
736 | blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); | |
817bf402 | 737 | put_dax(td->dm_dev.dax_dev); |
86f1152b | 738 | td->dm_dev.bdev = NULL; |
817bf402 | 739 | td->dm_dev.dax_dev = NULL; |
86f1152b BM |
740 | } |
741 | ||
742 | static struct table_device *find_table_device(struct list_head *l, dev_t dev, | |
743 | fmode_t mode) { | |
744 | struct table_device *td; | |
745 | ||
746 | list_for_each_entry(td, l, list) | |
747 | if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) | |
748 | return td; | |
749 | ||
750 | return NULL; | |
751 | } | |
752 | ||
753 | int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, | |
754 | struct dm_dev **result) { | |
755 | int r; | |
756 | struct table_device *td; | |
757 | ||
758 | mutex_lock(&md->table_devices_lock); | |
759 | td = find_table_device(&md->table_devices, dev, mode); | |
760 | if (!td) { | |
115485e8 | 761 | td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); |
86f1152b BM |
762 | if (!td) { |
763 | mutex_unlock(&md->table_devices_lock); | |
764 | return -ENOMEM; | |
765 | } | |
766 | ||
767 | td->dm_dev.mode = mode; | |
768 | td->dm_dev.bdev = NULL; | |
769 | ||
770 | if ((r = open_table_device(td, dev, md))) { | |
771 | mutex_unlock(&md->table_devices_lock); | |
772 | kfree(td); | |
773 | return r; | |
774 | } | |
775 | ||
776 | format_dev_t(td->dm_dev.name, dev); | |
777 | ||
b0b4d7c6 | 778 | refcount_set(&td->count, 1); |
86f1152b | 779 | list_add(&td->list, &md->table_devices); |
b0b4d7c6 ER |
780 | } else { |
781 | refcount_inc(&td->count); | |
86f1152b | 782 | } |
86f1152b BM |
783 | mutex_unlock(&md->table_devices_lock); |
784 | ||
785 | *result = &td->dm_dev; | |
786 | return 0; | |
787 | } | |
788 | EXPORT_SYMBOL_GPL(dm_get_table_device); | |
789 | ||
790 | void dm_put_table_device(struct mapped_device *md, struct dm_dev *d) | |
791 | { | |
792 | struct table_device *td = container_of(d, struct table_device, dm_dev); | |
793 | ||
794 | mutex_lock(&md->table_devices_lock); | |
b0b4d7c6 | 795 | if (refcount_dec_and_test(&td->count)) { |
86f1152b BM |
796 | close_table_device(td, md); |
797 | list_del(&td->list); | |
798 | kfree(td); | |
799 | } | |
800 | mutex_unlock(&md->table_devices_lock); | |
801 | } | |
802 | EXPORT_SYMBOL(dm_put_table_device); | |
803 | ||
804 | static void free_table_devices(struct list_head *devices) | |
805 | { | |
806 | struct list_head *tmp, *next; | |
807 | ||
808 | list_for_each_safe(tmp, next, devices) { | |
809 | struct table_device *td = list_entry(tmp, struct table_device, list); | |
810 | ||
811 | DMWARN("dm_destroy: %s still exists with %d references", | |
b0b4d7c6 | 812 | td->dm_dev.name, refcount_read(&td->count)); |
86f1152b BM |
813 | kfree(td); |
814 | } | |
815 | } | |
816 | ||
3ac51e74 DW |
817 | /* |
818 | * Get the geometry associated with a dm device | |
819 | */ | |
820 | int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
821 | { | |
822 | *geo = md->geometry; | |
823 | ||
824 | return 0; | |
825 | } | |
826 | ||
827 | /* | |
828 | * Set the geometry of a device. | |
829 | */ | |
830 | int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo) | |
831 | { | |
832 | sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors; | |
833 | ||
834 | if (geo->start > sz) { | |
835 | DMWARN("Start sector is beyond the geometry limits."); | |
836 | return -EINVAL; | |
837 | } | |
838 | ||
839 | md->geometry = *geo; | |
840 | ||
841 | return 0; | |
842 | } | |
843 | ||
2e93ccc1 KU |
844 | static int __noflush_suspending(struct mapped_device *md) |
845 | { | |
846 | return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
847 | } | |
848 | ||
1da177e4 LT |
849 | /* |
850 | * Decrements the number of outstanding ios that a bio has been | |
851 | * cloned into, completing the original io if necc. | |
852 | */ | |
4e4cbee9 | 853 | static void dec_pending(struct dm_io *io, blk_status_t error) |
1da177e4 | 854 | { |
2e93ccc1 | 855 | unsigned long flags; |
4e4cbee9 | 856 | blk_status_t io_error; |
b35f8caa MB |
857 | struct bio *bio; |
858 | struct mapped_device *md = io->md; | |
2e93ccc1 KU |
859 | |
860 | /* Push-back supersedes any I/O errors */ | |
f88fb981 KU |
861 | if (unlikely(error)) { |
862 | spin_lock_irqsave(&io->endio_lock, flags); | |
745dc570 | 863 | if (!(io->status == BLK_STS_DM_REQUEUE && __noflush_suspending(md))) |
4e4cbee9 | 864 | io->status = error; |
f88fb981 KU |
865 | spin_unlock_irqrestore(&io->endio_lock, flags); |
866 | } | |
1da177e4 LT |
867 | |
868 | if (atomic_dec_and_test(&io->io_count)) { | |
4e4cbee9 | 869 | if (io->status == BLK_STS_DM_REQUEUE) { |
2e93ccc1 KU |
870 | /* |
871 | * Target requested pushing back the I/O. | |
2e93ccc1 | 872 | */ |
022c2611 | 873 | spin_lock_irqsave(&md->deferred_lock, flags); |
6a8736d1 | 874 | if (__noflush_suspending(md)) |
745dc570 MS |
875 | /* NOTE early return due to BLK_STS_DM_REQUEUE below */ |
876 | bio_list_add_head(&md->deferred, io->orig_bio); | |
6a8736d1 | 877 | else |
2e93ccc1 | 878 | /* noflush suspend was interrupted. */ |
4e4cbee9 | 879 | io->status = BLK_STS_IOERR; |
022c2611 | 880 | spin_unlock_irqrestore(&md->deferred_lock, flags); |
2e93ccc1 KU |
881 | } |
882 | ||
4e4cbee9 | 883 | io_error = io->status; |
745dc570 | 884 | bio = io->orig_bio; |
6a8736d1 TH |
885 | end_io_acct(io); |
886 | free_io(md, io); | |
887 | ||
4e4cbee9 | 888 | if (io_error == BLK_STS_DM_REQUEUE) |
6a8736d1 | 889 | return; |
2e93ccc1 | 890 | |
1eff9d32 | 891 | if ((bio->bi_opf & REQ_PREFLUSH) && bio->bi_iter.bi_size) { |
af7e466a | 892 | /* |
6a8736d1 | 893 | * Preflush done for flush with data, reissue |
28a8f0d3 | 894 | * without REQ_PREFLUSH. |
af7e466a | 895 | */ |
1eff9d32 | 896 | bio->bi_opf &= ~REQ_PREFLUSH; |
6a8736d1 | 897 | queue_io(md, bio); |
af7e466a | 898 | } else { |
b372d360 | 899 | /* done with normal IO or empty flush */ |
4e4cbee9 | 900 | bio->bi_status = io_error; |
4246a0b6 | 901 | bio_endio(bio); |
b35f8caa | 902 | } |
1da177e4 LT |
903 | } |
904 | } | |
905 | ||
4cc96131 | 906 | void disable_write_same(struct mapped_device *md) |
7eee4ae2 MS |
907 | { |
908 | struct queue_limits *limits = dm_get_queue_limits(md); | |
909 | ||
910 | /* device doesn't really support WRITE SAME, disable it */ | |
911 | limits->max_write_same_sectors = 0; | |
912 | } | |
913 | ||
ac62d620 CH |
914 | void disable_write_zeroes(struct mapped_device *md) |
915 | { | |
916 | struct queue_limits *limits = dm_get_queue_limits(md); | |
917 | ||
918 | /* device doesn't really support WRITE ZEROES, disable it */ | |
919 | limits->max_write_zeroes_sectors = 0; | |
920 | } | |
921 | ||
4246a0b6 | 922 | static void clone_endio(struct bio *bio) |
1da177e4 | 923 | { |
4e4cbee9 | 924 | blk_status_t error = bio->bi_status; |
bfc6d41c | 925 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); |
b35f8caa | 926 | struct dm_io *io = tio->io; |
9faf400f | 927 | struct mapped_device *md = tio->io->md; |
1da177e4 LT |
928 | dm_endio_fn endio = tio->ti->type->end_io; |
929 | ||
4e4cbee9 | 930 | if (unlikely(error == BLK_STS_TARGET)) { |
ac62d620 | 931 | if (bio_op(bio) == REQ_OP_WRITE_SAME && |
74d46992 | 932 | !bio->bi_disk->queue->limits.max_write_same_sectors) |
ac62d620 CH |
933 | disable_write_same(md); |
934 | if (bio_op(bio) == REQ_OP_WRITE_ZEROES && | |
74d46992 | 935 | !bio->bi_disk->queue->limits.max_write_zeroes_sectors) |
ac62d620 CH |
936 | disable_write_zeroes(md); |
937 | } | |
7eee4ae2 | 938 | |
1be56909 | 939 | if (endio) { |
4e4cbee9 | 940 | int r = endio(tio->ti, bio, &error); |
1be56909 CH |
941 | switch (r) { |
942 | case DM_ENDIO_REQUEUE: | |
4e4cbee9 | 943 | error = BLK_STS_DM_REQUEUE; |
1be56909 CH |
944 | /*FALLTHRU*/ |
945 | case DM_ENDIO_DONE: | |
946 | break; | |
947 | case DM_ENDIO_INCOMPLETE: | |
948 | /* The target will handle the io */ | |
949 | return; | |
950 | default: | |
951 | DMWARN("unimplemented target endio return value: %d", r); | |
952 | BUG(); | |
953 | } | |
954 | } | |
955 | ||
cfae7529 | 956 | free_tio(tio); |
b35f8caa | 957 | dec_pending(io, error); |
1da177e4 LT |
958 | } |
959 | ||
56a67df7 MS |
960 | /* |
961 | * Return maximum size of I/O possible at the supplied sector up to the current | |
962 | * target boundary. | |
963 | */ | |
964 | static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti) | |
965 | { | |
966 | sector_t target_offset = dm_target_offset(ti, sector); | |
967 | ||
968 | return ti->len - target_offset; | |
969 | } | |
970 | ||
971 | static sector_t max_io_len(sector_t sector, struct dm_target *ti) | |
1da177e4 | 972 | { |
56a67df7 | 973 | sector_t len = max_io_len_target_boundary(sector, ti); |
542f9038 | 974 | sector_t offset, max_len; |
1da177e4 LT |
975 | |
976 | /* | |
542f9038 | 977 | * Does the target need to split even further? |
1da177e4 | 978 | */ |
542f9038 MS |
979 | if (ti->max_io_len) { |
980 | offset = dm_target_offset(ti, sector); | |
981 | if (unlikely(ti->max_io_len & (ti->max_io_len - 1))) | |
982 | max_len = sector_div(offset, ti->max_io_len); | |
983 | else | |
984 | max_len = offset & (ti->max_io_len - 1); | |
985 | max_len = ti->max_io_len - max_len; | |
986 | ||
987 | if (len > max_len) | |
988 | len = max_len; | |
1da177e4 LT |
989 | } |
990 | ||
991 | return len; | |
992 | } | |
993 | ||
542f9038 MS |
994 | int dm_set_target_max_io_len(struct dm_target *ti, sector_t len) |
995 | { | |
996 | if (len > UINT_MAX) { | |
997 | DMERR("Specified maximum size of target IO (%llu) exceeds limit (%u)", | |
998 | (unsigned long long)len, UINT_MAX); | |
999 | ti->error = "Maximum size of target IO is too large"; | |
1000 | return -EINVAL; | |
1001 | } | |
1002 | ||
1003 | ti->max_io_len = (uint32_t) len; | |
1004 | ||
1005 | return 0; | |
1006 | } | |
1007 | EXPORT_SYMBOL_GPL(dm_set_target_max_io_len); | |
1008 | ||
f26c5719 DW |
1009 | static struct dm_target *dm_dax_get_live_target(struct mapped_device *md, |
1010 | sector_t sector, int *srcu_idx) | |
545ed20e | 1011 | { |
545ed20e TK |
1012 | struct dm_table *map; |
1013 | struct dm_target *ti; | |
545ed20e | 1014 | |
f26c5719 | 1015 | map = dm_get_live_table(md, srcu_idx); |
545ed20e | 1016 | if (!map) |
f26c5719 | 1017 | return NULL; |
545ed20e TK |
1018 | |
1019 | ti = dm_table_find_target(map, sector); | |
1020 | if (!dm_target_is_valid(ti)) | |
f26c5719 | 1021 | return NULL; |
545ed20e | 1022 | |
f26c5719 DW |
1023 | return ti; |
1024 | } | |
545ed20e | 1025 | |
f26c5719 DW |
1026 | static long dm_dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, |
1027 | long nr_pages, void **kaddr, pfn_t *pfn) | |
1028 | { | |
1029 | struct mapped_device *md = dax_get_private(dax_dev); | |
1030 | sector_t sector = pgoff * PAGE_SECTORS; | |
1031 | struct dm_target *ti; | |
1032 | long len, ret = -EIO; | |
1033 | int srcu_idx; | |
545ed20e | 1034 | |
f26c5719 | 1035 | ti = dm_dax_get_live_target(md, sector, &srcu_idx); |
545ed20e | 1036 | |
f26c5719 DW |
1037 | if (!ti) |
1038 | goto out; | |
1039 | if (!ti->type->direct_access) | |
1040 | goto out; | |
1041 | len = max_io_len(sector, ti) / PAGE_SECTORS; | |
1042 | if (len < 1) | |
1043 | goto out; | |
1044 | nr_pages = min(len, nr_pages); | |
545ed20e | 1045 | if (ti->type->direct_access) |
817bf402 DW |
1046 | ret = ti->type->direct_access(ti, pgoff, nr_pages, kaddr, pfn); |
1047 | ||
f26c5719 | 1048 | out: |
545ed20e | 1049 | dm_put_live_table(md, srcu_idx); |
f26c5719 DW |
1050 | |
1051 | return ret; | |
545ed20e TK |
1052 | } |
1053 | ||
7e026c8c DW |
1054 | static size_t dm_dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, |
1055 | void *addr, size_t bytes, struct iov_iter *i) | |
1056 | { | |
1057 | struct mapped_device *md = dax_get_private(dax_dev); | |
1058 | sector_t sector = pgoff * PAGE_SECTORS; | |
1059 | struct dm_target *ti; | |
1060 | long ret = 0; | |
1061 | int srcu_idx; | |
1062 | ||
1063 | ti = dm_dax_get_live_target(md, sector, &srcu_idx); | |
1064 | ||
1065 | if (!ti) | |
1066 | goto out; | |
1067 | if (!ti->type->dax_copy_from_iter) { | |
1068 | ret = copy_from_iter(addr, bytes, i); | |
1069 | goto out; | |
1070 | } | |
1071 | ret = ti->type->dax_copy_from_iter(ti, pgoff, addr, bytes, i); | |
1072 | out: | |
1073 | dm_put_live_table(md, srcu_idx); | |
1074 | ||
1075 | return ret; | |
1076 | } | |
1077 | ||
1dd40c3e MP |
1078 | /* |
1079 | * A target may call dm_accept_partial_bio only from the map routine. It is | |
c06b3e58 | 1080 | * allowed for all bio types except REQ_PREFLUSH and REQ_OP_ZONE_RESET. |
1dd40c3e MP |
1081 | * |
1082 | * dm_accept_partial_bio informs the dm that the target only wants to process | |
1083 | * additional n_sectors sectors of the bio and the rest of the data should be | |
1084 | * sent in a next bio. | |
1085 | * | |
1086 | * A diagram that explains the arithmetics: | |
1087 | * +--------------------+---------------+-------+ | |
1088 | * | 1 | 2 | 3 | | |
1089 | * +--------------------+---------------+-------+ | |
1090 | * | |
1091 | * <-------------- *tio->len_ptr ---------------> | |
1092 | * <------- bi_size -------> | |
1093 | * <-- n_sectors --> | |
1094 | * | |
1095 | * Region 1 was already iterated over with bio_advance or similar function. | |
1096 | * (it may be empty if the target doesn't use bio_advance) | |
1097 | * Region 2 is the remaining bio size that the target wants to process. | |
1098 | * (it may be empty if region 1 is non-empty, although there is no reason | |
1099 | * to make it empty) | |
1100 | * The target requires that region 3 is to be sent in the next bio. | |
1101 | * | |
1102 | * If the target wants to receive multiple copies of the bio (via num_*bios, etc), | |
1103 | * the partially processed part (the sum of regions 1+2) must be the same for all | |
1104 | * copies of the bio. | |
1105 | */ | |
1106 | void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors) | |
1107 | { | |
1108 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); | |
1109 | unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT; | |
1eff9d32 | 1110 | BUG_ON(bio->bi_opf & REQ_PREFLUSH); |
1dd40c3e MP |
1111 | BUG_ON(bi_size > *tio->len_ptr); |
1112 | BUG_ON(n_sectors > bi_size); | |
1113 | *tio->len_ptr -= bi_size - n_sectors; | |
1114 | bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT; | |
1115 | } | |
1116 | EXPORT_SYMBOL_GPL(dm_accept_partial_bio); | |
1117 | ||
10999307 DLM |
1118 | /* |
1119 | * The zone descriptors obtained with a zone report indicate | |
1120 | * zone positions within the target device. The zone descriptors | |
1121 | * must be remapped to match their position within the dm device. | |
1122 | * A target may call dm_remap_zone_report after completion of a | |
1123 | * REQ_OP_ZONE_REPORT bio to remap the zone descriptors obtained | |
1124 | * from the target device mapping to the dm device. | |
1125 | */ | |
1126 | void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, sector_t start) | |
1127 | { | |
1128 | #ifdef CONFIG_BLK_DEV_ZONED | |
1129 | struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone); | |
745dc570 | 1130 | struct bio *report_bio = tio->io->orig_bio; |
10999307 DLM |
1131 | struct blk_zone_report_hdr *hdr = NULL; |
1132 | struct blk_zone *zone; | |
1133 | unsigned int nr_rep = 0; | |
1134 | unsigned int ofst; | |
1135 | struct bio_vec bvec; | |
1136 | struct bvec_iter iter; | |
1137 | void *addr; | |
1138 | ||
1139 | if (bio->bi_status) | |
1140 | return; | |
1141 | ||
1142 | /* | |
1143 | * Remap the start sector of the reported zones. For sequential zones, | |
1144 | * also remap the write pointer position. | |
1145 | */ | |
1146 | bio_for_each_segment(bvec, report_bio, iter) { | |
1147 | addr = kmap_atomic(bvec.bv_page); | |
1148 | ||
1149 | /* Remember the report header in the first page */ | |
1150 | if (!hdr) { | |
1151 | hdr = addr; | |
1152 | ofst = sizeof(struct blk_zone_report_hdr); | |
1153 | } else | |
1154 | ofst = 0; | |
1155 | ||
1156 | /* Set zones start sector */ | |
1157 | while (hdr->nr_zones && ofst < bvec.bv_len) { | |
1158 | zone = addr + ofst; | |
1159 | if (zone->start >= start + ti->len) { | |
1160 | hdr->nr_zones = 0; | |
1161 | break; | |
1162 | } | |
1163 | zone->start = zone->start + ti->begin - start; | |
1164 | if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) { | |
1165 | if (zone->cond == BLK_ZONE_COND_FULL) | |
1166 | zone->wp = zone->start + zone->len; | |
1167 | else if (zone->cond == BLK_ZONE_COND_EMPTY) | |
1168 | zone->wp = zone->start; | |
1169 | else | |
1170 | zone->wp = zone->wp + ti->begin - start; | |
1171 | } | |
1172 | ofst += sizeof(struct blk_zone); | |
1173 | hdr->nr_zones--; | |
1174 | nr_rep++; | |
1175 | } | |
1176 | ||
1177 | if (addr != hdr) | |
1178 | kunmap_atomic(addr); | |
1179 | ||
1180 | if (!hdr->nr_zones) | |
1181 | break; | |
1182 | } | |
1183 | ||
1184 | if (hdr) { | |
1185 | hdr->nr_zones = nr_rep; | |
1186 | kunmap_atomic(hdr); | |
1187 | } | |
1188 | ||
1189 | bio_advance(report_bio, report_bio->bi_iter.bi_size); | |
1190 | ||
1191 | #else /* !CONFIG_BLK_DEV_ZONED */ | |
1192 | bio->bi_status = BLK_STS_NOTSUPP; | |
1193 | #endif | |
1194 | } | |
1195 | EXPORT_SYMBOL_GPL(dm_remap_zone_report); | |
1196 | ||
bd2a49b8 | 1197 | static void __map_bio(struct dm_target_io *tio) |
1da177e4 LT |
1198 | { |
1199 | int r; | |
2056a782 | 1200 | sector_t sector; |
dba14160 | 1201 | struct bio *clone = &tio->clone; |
64f52b0e | 1202 | struct dm_io *io = tio->io; |
bd2a49b8 | 1203 | struct dm_target *ti = tio->ti; |
1da177e4 | 1204 | |
1da177e4 | 1205 | clone->bi_end_io = clone_endio; |
1da177e4 LT |
1206 | |
1207 | /* | |
1208 | * Map the clone. If r == 0 we don't need to do | |
1209 | * anything, the target has assumed ownership of | |
1210 | * this io. | |
1211 | */ | |
64f52b0e | 1212 | atomic_inc(&io->io_count); |
4f024f37 | 1213 | sector = clone->bi_iter.bi_sector; |
d67a5f4b | 1214 | |
7de3ee57 | 1215 | r = ti->type->map(ti, clone); |
846785e6 CH |
1216 | switch (r) { |
1217 | case DM_MAPIO_SUBMITTED: | |
1218 | break; | |
1219 | case DM_MAPIO_REMAPPED: | |
1da177e4 | 1220 | /* the bio has been remapped so dispatch it */ |
74d46992 | 1221 | trace_block_bio_remap(clone->bi_disk->queue, clone, |
64f52b0e | 1222 | bio_dev(io->orig_bio), sector); |
1da177e4 | 1223 | generic_make_request(clone); |
846785e6 CH |
1224 | break; |
1225 | case DM_MAPIO_KILL: | |
4e4cbee9 | 1226 | free_tio(tio); |
64f52b0e | 1227 | dec_pending(io, BLK_STS_IOERR); |
4e4cbee9 | 1228 | break; |
846785e6 | 1229 | case DM_MAPIO_REQUEUE: |
cfae7529 | 1230 | free_tio(tio); |
64f52b0e | 1231 | dec_pending(io, BLK_STS_DM_REQUEUE); |
846785e6 CH |
1232 | break; |
1233 | default: | |
45cbcd79 KU |
1234 | DMWARN("unimplemented target map return value: %d", r); |
1235 | BUG(); | |
1da177e4 LT |
1236 | } |
1237 | } | |
1238 | ||
e0d6609a | 1239 | static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len) |
bd2a49b8 | 1240 | { |
4f024f37 KO |
1241 | bio->bi_iter.bi_sector = sector; |
1242 | bio->bi_iter.bi_size = to_bytes(len); | |
1da177e4 LT |
1243 | } |
1244 | ||
1245 | /* | |
1246 | * Creates a bio that consists of range of complete bvecs. | |
1247 | */ | |
c80914e8 MS |
1248 | static int clone_bio(struct dm_target_io *tio, struct bio *bio, |
1249 | sector_t sector, unsigned len) | |
1da177e4 | 1250 | { |
dba14160 | 1251 | struct bio *clone = &tio->clone; |
1da177e4 | 1252 | |
1c3b13e6 KO |
1253 | __bio_clone_fast(clone, bio); |
1254 | ||
e2460f2a MP |
1255 | if (unlikely(bio_integrity(bio) != NULL)) { |
1256 | int r; | |
1257 | ||
1258 | if (unlikely(!dm_target_has_integrity(tio->ti->type) && | |
1259 | !dm_target_passes_integrity(tio->ti->type))) { | |
1260 | DMWARN("%s: the target %s doesn't support integrity data.", | |
1261 | dm_device_name(tio->io->md), | |
1262 | tio->ti->type->name); | |
1263 | return -EIO; | |
1264 | } | |
1265 | ||
1266 | r = bio_integrity_clone(clone, bio, GFP_NOIO); | |
c80914e8 MS |
1267 | if (r < 0) |
1268 | return r; | |
1269 | } | |
bd2a49b8 | 1270 | |
264c869d DLM |
1271 | if (bio_op(bio) != REQ_OP_ZONE_REPORT) |
1272 | bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector)); | |
1c3b13e6 KO |
1273 | clone->bi_iter.bi_size = to_bytes(len); |
1274 | ||
e2460f2a | 1275 | if (unlikely(bio_integrity(bio) != NULL)) |
fbd08e76 | 1276 | bio_integrity_trim(clone); |
c80914e8 MS |
1277 | |
1278 | return 0; | |
1da177e4 LT |
1279 | } |
1280 | ||
318716dd MS |
1281 | static void alloc_multiple_bios(struct bio_list *blist, struct clone_info *ci, |
1282 | struct dm_target *ti, unsigned num_bios) | |
1283 | { | |
1284 | struct dm_target_io *tio; | |
1285 | int try; | |
1286 | ||
1287 | if (!num_bios) | |
1288 | return; | |
1289 | ||
1290 | if (num_bios == 1) { | |
1291 | tio = alloc_tio(ci, ti, 0, GFP_NOIO); | |
1292 | bio_list_add(blist, &tio->clone); | |
1293 | return; | |
1294 | } | |
1295 | ||
1296 | for (try = 0; try < 2; try++) { | |
1297 | int bio_nr; | |
1298 | struct bio *bio; | |
1299 | ||
1300 | if (try) | |
1301 | mutex_lock(&ci->md->table_devices_lock); | |
1302 | for (bio_nr = 0; bio_nr < num_bios; bio_nr++) { | |
1303 | tio = alloc_tio(ci, ti, bio_nr, try ? GFP_NOIO : GFP_NOWAIT); | |
1304 | if (!tio) | |
1305 | break; | |
1306 | ||
1307 | bio_list_add(blist, &tio->clone); | |
1308 | } | |
1309 | if (try) | |
1310 | mutex_unlock(&ci->md->table_devices_lock); | |
1311 | if (bio_nr == num_bios) | |
1312 | return; | |
1313 | ||
1314 | while ((bio = bio_list_pop(blist))) { | |
1315 | tio = container_of(bio, struct dm_target_io, clone); | |
1316 | free_tio(tio); | |
1317 | } | |
1318 | } | |
1319 | } | |
1320 | ||
14fe594d | 1321 | static void __clone_and_map_simple_bio(struct clone_info *ci, |
318716dd | 1322 | struct dm_target_io *tio, unsigned *len) |
9015df24 | 1323 | { |
dba14160 | 1324 | struct bio *clone = &tio->clone; |
9015df24 | 1325 | |
1dd40c3e MP |
1326 | tio->len_ptr = len; |
1327 | ||
99778273 | 1328 | __bio_clone_fast(clone, ci->bio); |
bd2a49b8 | 1329 | if (len) |
1dd40c3e | 1330 | bio_setup_sector(clone, ci->sector, *len); |
f9ab94ce | 1331 | |
bd2a49b8 | 1332 | __map_bio(tio); |
f9ab94ce MP |
1333 | } |
1334 | ||
14fe594d | 1335 | static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti, |
1dd40c3e | 1336 | unsigned num_bios, unsigned *len) |
06a426ce | 1337 | { |
318716dd MS |
1338 | struct bio_list blist = BIO_EMPTY_LIST; |
1339 | struct bio *bio; | |
1340 | struct dm_target_io *tio; | |
1341 | ||
1342 | alloc_multiple_bios(&blist, ci, ti, num_bios); | |
06a426ce | 1343 | |
318716dd MS |
1344 | while ((bio = bio_list_pop(&blist))) { |
1345 | tio = container_of(bio, struct dm_target_io, clone); | |
1346 | __clone_and_map_simple_bio(ci, tio, len); | |
1347 | } | |
06a426ce MS |
1348 | } |
1349 | ||
14fe594d | 1350 | static int __send_empty_flush(struct clone_info *ci) |
f9ab94ce | 1351 | { |
06a426ce | 1352 | unsigned target_nr = 0; |
f9ab94ce MP |
1353 | struct dm_target *ti; |
1354 | ||
b372d360 | 1355 | BUG_ON(bio_has_data(ci->bio)); |
f9ab94ce | 1356 | while ((ti = dm_table_get_target(ci->map, target_nr++))) |
1dd40c3e | 1357 | __send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL); |
f9ab94ce | 1358 | |
f9ab94ce MP |
1359 | return 0; |
1360 | } | |
1361 | ||
c80914e8 | 1362 | static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti, |
f31c21e4 | 1363 | sector_t sector, unsigned *len) |
5ae89a87 | 1364 | { |
dba14160 | 1365 | struct bio *bio = ci->bio; |
5ae89a87 | 1366 | struct dm_target_io *tio; |
f31c21e4 | 1367 | int r; |
5ae89a87 | 1368 | |
318716dd | 1369 | tio = alloc_tio(ci, ti, 0, GFP_NOIO); |
f31c21e4 N |
1370 | tio->len_ptr = len; |
1371 | r = clone_bio(tio, bio, sector, *len); | |
1372 | if (r < 0) { | |
1373 | free_tio(tio); | |
1374 | return r; | |
b0d8ed4d | 1375 | } |
f31c21e4 | 1376 | __map_bio(tio); |
c80914e8 | 1377 | |
f31c21e4 | 1378 | return 0; |
5ae89a87 MS |
1379 | } |
1380 | ||
55a62eef | 1381 | typedef unsigned (*get_num_bios_fn)(struct dm_target *ti); |
23508a96 | 1382 | |
55a62eef | 1383 | static unsigned get_num_discard_bios(struct dm_target *ti) |
23508a96 | 1384 | { |
55a62eef | 1385 | return ti->num_discard_bios; |
23508a96 MS |
1386 | } |
1387 | ||
55a62eef | 1388 | static unsigned get_num_write_same_bios(struct dm_target *ti) |
23508a96 | 1389 | { |
55a62eef | 1390 | return ti->num_write_same_bios; |
23508a96 MS |
1391 | } |
1392 | ||
ac62d620 CH |
1393 | static unsigned get_num_write_zeroes_bios(struct dm_target *ti) |
1394 | { | |
1395 | return ti->num_write_zeroes_bios; | |
1396 | } | |
1397 | ||
23508a96 | 1398 | typedef bool (*is_split_required_fn)(struct dm_target *ti); |
9eef87da | 1399 | |
23508a96 MS |
1400 | static bool is_split_required_for_discard(struct dm_target *ti) |
1401 | { | |
55a62eef | 1402 | return ti->split_discard_bios; |
cec47e3d KU |
1403 | } |
1404 | ||
3d7f4562 | 1405 | static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *ti, |
14fe594d AK |
1406 | get_num_bios_fn get_num_bios, |
1407 | is_split_required_fn is_split_required) | |
ba1cbad9 | 1408 | { |
e0d6609a | 1409 | unsigned len; |
55a62eef | 1410 | unsigned num_bios; |
ba1cbad9 | 1411 | |
3d7f4562 MS |
1412 | /* |
1413 | * Even though the device advertised support for this type of | |
1414 | * request, that does not mean every target supports it, and | |
1415 | * reconfiguration might also have changed that since the | |
1416 | * check was performed. | |
1417 | */ | |
1418 | num_bios = get_num_bios ? get_num_bios(ti) : 0; | |
1419 | if (!num_bios) | |
1420 | return -EOPNOTSUPP; | |
ba1cbad9 | 1421 | |
3d7f4562 MS |
1422 | if (is_split_required && !is_split_required(ti)) |
1423 | len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti)); | |
1424 | else | |
1425 | len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti)); | |
de3ec86d | 1426 | |
3d7f4562 | 1427 | __send_duplicate_bios(ci, ti, num_bios, &len); |
e262f347 | 1428 | |
3d7f4562 MS |
1429 | ci->sector += len; |
1430 | ci->sector_count -= len; | |
5ae89a87 MS |
1431 | |
1432 | return 0; | |
ba1cbad9 MS |
1433 | } |
1434 | ||
3d7f4562 | 1435 | static int __send_discard(struct clone_info *ci, struct dm_target *ti) |
23508a96 | 1436 | { |
3d7f4562 | 1437 | return __send_changing_extent_only(ci, ti, get_num_discard_bios, |
14fe594d | 1438 | is_split_required_for_discard); |
23508a96 | 1439 | } |
0ce65797 | 1440 | |
3d7f4562 | 1441 | static int __send_write_same(struct clone_info *ci, struct dm_target *ti) |
0ce65797 | 1442 | { |
3d7f4562 | 1443 | return __send_changing_extent_only(ci, ti, get_num_write_same_bios, NULL); |
0ce65797 MS |
1444 | } |
1445 | ||
3d7f4562 | 1446 | static int __send_write_zeroes(struct clone_info *ci, struct dm_target *ti) |
ac62d620 | 1447 | { |
3d7f4562 | 1448 | return __send_changing_extent_only(ci, ti, get_num_write_zeroes_bios, NULL); |
ac62d620 CH |
1449 | } |
1450 | ||
e4c93811 AK |
1451 | /* |
1452 | * Select the correct strategy for processing a non-flush bio. | |
1453 | */ | |
14fe594d | 1454 | static int __split_and_process_non_flush(struct clone_info *ci) |
0ce65797 | 1455 | { |
dba14160 | 1456 | struct bio *bio = ci->bio; |
512875bd | 1457 | struct dm_target *ti; |
1c3b13e6 | 1458 | unsigned len; |
c80914e8 | 1459 | int r; |
0ce65797 | 1460 | |
512875bd JN |
1461 | ti = dm_table_find_target(ci->map, ci->sector); |
1462 | if (!dm_target_is_valid(ti)) | |
1463 | return -EIO; | |
1464 | ||
3d7f4562 MS |
1465 | if (unlikely(bio_op(bio) == REQ_OP_DISCARD)) |
1466 | return __send_discard(ci, ti); | |
1467 | else if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) | |
1468 | return __send_write_same(ci, ti); | |
1469 | else if (unlikely(bio_op(bio) == REQ_OP_WRITE_ZEROES)) | |
1470 | return __send_write_zeroes(ci, ti); | |
1471 | ||
264c869d DLM |
1472 | if (bio_op(bio) == REQ_OP_ZONE_REPORT) |
1473 | len = ci->sector_count; | |
1474 | else | |
1475 | len = min_t(sector_t, max_io_len(ci->sector, ti), | |
1476 | ci->sector_count); | |
0ce65797 | 1477 | |
c80914e8 MS |
1478 | r = __clone_and_map_data_bio(ci, ti, ci->sector, &len); |
1479 | if (r < 0) | |
1480 | return r; | |
0ce65797 | 1481 | |
1c3b13e6 KO |
1482 | ci->sector += len; |
1483 | ci->sector_count -= len; | |
0ce65797 | 1484 | |
1c3b13e6 | 1485 | return 0; |
0ce65797 MS |
1486 | } |
1487 | ||
1da177e4 | 1488 | /* |
14fe594d | 1489 | * Entry point to split a bio into clones and submit them to the targets. |
1da177e4 | 1490 | */ |
83d5e5b0 MP |
1491 | static void __split_and_process_bio(struct mapped_device *md, |
1492 | struct dm_table *map, struct bio *bio) | |
0ce65797 | 1493 | { |
1da177e4 | 1494 | struct clone_info ci; |
512875bd | 1495 | int error = 0; |
1da177e4 | 1496 | |
83d5e5b0 | 1497 | if (unlikely(!map)) { |
6a8736d1 | 1498 | bio_io_error(bio); |
f0b9a450 MP |
1499 | return; |
1500 | } | |
692d0eb9 | 1501 | |
83d5e5b0 | 1502 | ci.map = map; |
1da177e4 | 1503 | ci.md = md; |
1da177e4 | 1504 | ci.io = alloc_io(md); |
4e4cbee9 | 1505 | ci.io->status = 0; |
1da177e4 | 1506 | atomic_set(&ci.io->io_count, 1); |
745dc570 | 1507 | ci.io->orig_bio = bio; |
1da177e4 | 1508 | ci.io->md = md; |
f88fb981 | 1509 | spin_lock_init(&ci.io->endio_lock); |
4f024f37 | 1510 | ci.sector = bio->bi_iter.bi_sector; |
0ce65797 | 1511 | |
3eaf840e | 1512 | start_io_acct(ci.io); |
0ce65797 | 1513 | |
1eff9d32 | 1514 | if (bio->bi_opf & REQ_PREFLUSH) { |
b372d360 MS |
1515 | ci.bio = &ci.md->flush_bio; |
1516 | ci.sector_count = 0; | |
14fe594d | 1517 | error = __send_empty_flush(&ci); |
b372d360 | 1518 | /* dec_pending submits any data associated with flush */ |
a4aa5e56 DLM |
1519 | } else if (bio_op(bio) == REQ_OP_ZONE_RESET) { |
1520 | ci.bio = bio; | |
1521 | ci.sector_count = 0; | |
1522 | error = __split_and_process_non_flush(&ci); | |
b372d360 | 1523 | } else { |
6a8736d1 | 1524 | ci.bio = bio; |
d87f4c14 | 1525 | ci.sector_count = bio_sectors(bio); |
18a25da8 | 1526 | while (ci.sector_count && !error) { |
14fe594d | 1527 | error = __split_and_process_non_flush(&ci); |
18a25da8 N |
1528 | if (current->bio_list && ci.sector_count && !error) { |
1529 | /* | |
1530 | * Remainder must be passed to generic_make_request() | |
1531 | * so that it gets handled *after* bios already submitted | |
1532 | * have been completely processed. | |
1533 | * We take a clone of the original to store in | |
745dc570 | 1534 | * ci.io->orig_bio to be used by end_io_acct() and |
18a25da8 N |
1535 | * for dec_pending to use for completion handling. |
1536 | * As this path is not used for REQ_OP_ZONE_REPORT, | |
745dc570 | 1537 | * the usage of io->orig_bio in dm_remap_zone_report() |
18a25da8 N |
1538 | * won't be affected by this reassignment. |
1539 | */ | |
1540 | struct bio *b = bio_clone_bioset(bio, GFP_NOIO, | |
1541 | md->queue->bio_split); | |
745dc570 | 1542 | ci.io->orig_bio = b; |
18a25da8 N |
1543 | bio_advance(bio, (bio_sectors(bio) - ci.sector_count) << 9); |
1544 | bio_chain(b, bio); | |
1545 | generic_make_request(bio); | |
1546 | break; | |
1547 | } | |
1548 | } | |
d87f4c14 | 1549 | } |
0ce65797 | 1550 | |
1da177e4 | 1551 | /* drop the extra reference count */ |
54385bf7 | 1552 | dec_pending(ci.io, errno_to_blk_status(error)); |
0ce65797 MS |
1553 | } |
1554 | ||
cec47e3d | 1555 | /* |
18a25da8 N |
1556 | * The request function that remaps the bio to one target and |
1557 | * splits off any remainder. | |
cec47e3d | 1558 | */ |
dece1635 | 1559 | static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio) |
cec47e3d | 1560 | { |
12f03a49 | 1561 | int rw = bio_data_dir(bio); |
cec47e3d | 1562 | struct mapped_device *md = q->queuedata; |
83d5e5b0 MP |
1563 | int srcu_idx; |
1564 | struct dm_table *map; | |
cec47e3d | 1565 | |
83d5e5b0 | 1566 | map = dm_get_live_table(md, &srcu_idx); |
29e4013d | 1567 | |
d62e26b3 | 1568 | generic_start_io_acct(q, rw, bio_sectors(bio), &dm_disk(md)->part0); |
d0bcb878 | 1569 | |
6a8736d1 TH |
1570 | /* if we're suspended, we have to queue this io for later */ |
1571 | if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) { | |
83d5e5b0 | 1572 | dm_put_live_table(md, srcu_idx); |
9eef87da | 1573 | |
1eff9d32 | 1574 | if (!(bio->bi_opf & REQ_RAHEAD)) |
6a8736d1 TH |
1575 | queue_io(md, bio); |
1576 | else | |
54d9a1b4 | 1577 | bio_io_error(bio); |
dece1635 | 1578 | return BLK_QC_T_NONE; |
cec47e3d | 1579 | } |
1da177e4 | 1580 | |
83d5e5b0 MP |
1581 | __split_and_process_bio(md, map, bio); |
1582 | dm_put_live_table(md, srcu_idx); | |
dece1635 | 1583 | return BLK_QC_T_NONE; |
cec47e3d KU |
1584 | } |
1585 | ||
1da177e4 LT |
1586 | static int dm_any_congested(void *congested_data, int bdi_bits) |
1587 | { | |
8a57dfc6 CS |
1588 | int r = bdi_bits; |
1589 | struct mapped_device *md = congested_data; | |
1590 | struct dm_table *map; | |
1da177e4 | 1591 | |
1eb787ec | 1592 | if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
e522c039 | 1593 | if (dm_request_based(md)) { |
cec47e3d | 1594 | /* |
e522c039 MS |
1595 | * With request-based DM we only need to check the |
1596 | * top-level queue for congestion. | |
cec47e3d | 1597 | */ |
dc3b17cc | 1598 | r = md->queue->backing_dev_info->wb.state & bdi_bits; |
e522c039 MS |
1599 | } else { |
1600 | map = dm_get_live_table_fast(md); | |
1601 | if (map) | |
cec47e3d | 1602 | r = dm_table_any_congested(map, bdi_bits); |
e522c039 | 1603 | dm_put_live_table_fast(md); |
8a57dfc6 CS |
1604 | } |
1605 | } | |
1606 | ||
1da177e4 LT |
1607 | return r; |
1608 | } | |
1609 | ||
1610 | /*----------------------------------------------------------------- | |
1611 | * An IDR is used to keep track of allocated minor numbers. | |
1612 | *---------------------------------------------------------------*/ | |
2b06cfff | 1613 | static void free_minor(int minor) |
1da177e4 | 1614 | { |
f32c10b0 | 1615 | spin_lock(&_minor_lock); |
1da177e4 | 1616 | idr_remove(&_minor_idr, minor); |
f32c10b0 | 1617 | spin_unlock(&_minor_lock); |
1da177e4 LT |
1618 | } |
1619 | ||
1620 | /* | |
1621 | * See if the device with a specific minor # is free. | |
1622 | */ | |
cf13ab8e | 1623 | static int specific_minor(int minor) |
1da177e4 | 1624 | { |
c9d76be6 | 1625 | int r; |
1da177e4 LT |
1626 | |
1627 | if (minor >= (1 << MINORBITS)) | |
1628 | return -EINVAL; | |
1629 | ||
c9d76be6 | 1630 | idr_preload(GFP_KERNEL); |
f32c10b0 | 1631 | spin_lock(&_minor_lock); |
1da177e4 | 1632 | |
c9d76be6 | 1633 | r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT); |
1da177e4 | 1634 | |
f32c10b0 | 1635 | spin_unlock(&_minor_lock); |
c9d76be6 TH |
1636 | idr_preload_end(); |
1637 | if (r < 0) | |
1638 | return r == -ENOSPC ? -EBUSY : r; | |
1639 | return 0; | |
1da177e4 LT |
1640 | } |
1641 | ||
cf13ab8e | 1642 | static int next_free_minor(int *minor) |
1da177e4 | 1643 | { |
c9d76be6 | 1644 | int r; |
62f75c2f | 1645 | |
c9d76be6 | 1646 | idr_preload(GFP_KERNEL); |
f32c10b0 | 1647 | spin_lock(&_minor_lock); |
1da177e4 | 1648 | |
c9d76be6 | 1649 | r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT); |
1da177e4 | 1650 | |
f32c10b0 | 1651 | spin_unlock(&_minor_lock); |
c9d76be6 TH |
1652 | idr_preload_end(); |
1653 | if (r < 0) | |
1654 | return r; | |
1655 | *minor = r; | |
1656 | return 0; | |
1da177e4 LT |
1657 | } |
1658 | ||
83d5cde4 | 1659 | static const struct block_device_operations dm_blk_dops; |
f26c5719 | 1660 | static const struct dax_operations dm_dax_ops; |
1da177e4 | 1661 | |
53d5914f MP |
1662 | static void dm_wq_work(struct work_struct *work); |
1663 | ||
4cc96131 | 1664 | void dm_init_md_queue(struct mapped_device *md) |
4a0b4ddf | 1665 | { |
ad5f498f MP |
1666 | /* |
1667 | * Initialize data that will only be used by a non-blk-mq DM queue | |
1668 | * - must do so here (in alloc_dev callchain) before queue is used | |
1669 | */ | |
1670 | md->queue->queuedata = md; | |
dc3b17cc | 1671 | md->queue->backing_dev_info->congested_data = md; |
bfebd1cd | 1672 | } |
4a0b4ddf | 1673 | |
4cc96131 | 1674 | void dm_init_normal_md_queue(struct mapped_device *md) |
bfebd1cd | 1675 | { |
17e149b8 | 1676 | md->use_blk_mq = false; |
bfebd1cd MS |
1677 | dm_init_md_queue(md); |
1678 | ||
1679 | /* | |
1680 | * Initialize aspects of queue that aren't relevant for blk-mq | |
1681 | */ | |
dc3b17cc | 1682 | md->queue->backing_dev_info->congested_fn = dm_any_congested; |
4a0b4ddf MS |
1683 | } |
1684 | ||
0f20972f MS |
1685 | static void cleanup_mapped_device(struct mapped_device *md) |
1686 | { | |
0f20972f MS |
1687 | if (md->wq) |
1688 | destroy_workqueue(md->wq); | |
1689 | if (md->kworker_task) | |
1690 | kthread_stop(md->kworker_task); | |
0f20972f MS |
1691 | if (md->bs) |
1692 | bioset_free(md->bs); | |
64f52b0e MS |
1693 | if (md->io_bs) |
1694 | bioset_free(md->io_bs); | |
0f20972f | 1695 | |
f26c5719 DW |
1696 | if (md->dax_dev) { |
1697 | kill_dax(md->dax_dev); | |
1698 | put_dax(md->dax_dev); | |
1699 | md->dax_dev = NULL; | |
1700 | } | |
1701 | ||
0f20972f MS |
1702 | if (md->disk) { |
1703 | spin_lock(&_minor_lock); | |
1704 | md->disk->private_data = NULL; | |
1705 | spin_unlock(&_minor_lock); | |
0f20972f MS |
1706 | del_gendisk(md->disk); |
1707 | put_disk(md->disk); | |
1708 | } | |
1709 | ||
1710 | if (md->queue) | |
1711 | blk_cleanup_queue(md->queue); | |
1712 | ||
d09960b0 TE |
1713 | cleanup_srcu_struct(&md->io_barrier); |
1714 | ||
0f20972f MS |
1715 | if (md->bdev) { |
1716 | bdput(md->bdev); | |
1717 | md->bdev = NULL; | |
1718 | } | |
4cc96131 MS |
1719 | |
1720 | dm_mq_cleanup_mapped_device(md); | |
0f20972f MS |
1721 | } |
1722 | ||
1da177e4 LT |
1723 | /* |
1724 | * Allocate and initialise a blank device with a given minor. | |
1725 | */ | |
2b06cfff | 1726 | static struct mapped_device *alloc_dev(int minor) |
1da177e4 | 1727 | { |
115485e8 | 1728 | int r, numa_node_id = dm_get_numa_node(); |
f26c5719 | 1729 | struct dax_device *dax_dev; |
115485e8 | 1730 | struct mapped_device *md; |
ba61fdd1 | 1731 | void *old_md; |
1da177e4 | 1732 | |
856eb091 | 1733 | md = kvzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id); |
1da177e4 LT |
1734 | if (!md) { |
1735 | DMWARN("unable to allocate device, out of memory."); | |
1736 | return NULL; | |
1737 | } | |
1738 | ||
10da4f79 | 1739 | if (!try_module_get(THIS_MODULE)) |
6ed7ade8 | 1740 | goto bad_module_get; |
10da4f79 | 1741 | |
1da177e4 | 1742 | /* get a minor number for the dev */ |
2b06cfff | 1743 | if (minor == DM_ANY_MINOR) |
cf13ab8e | 1744 | r = next_free_minor(&minor); |
2b06cfff | 1745 | else |
cf13ab8e | 1746 | r = specific_minor(minor); |
1da177e4 | 1747 | if (r < 0) |
6ed7ade8 | 1748 | goto bad_minor; |
1da177e4 | 1749 | |
83d5e5b0 MP |
1750 | r = init_srcu_struct(&md->io_barrier); |
1751 | if (r < 0) | |
1752 | goto bad_io_barrier; | |
1753 | ||
115485e8 | 1754 | md->numa_node_id = numa_node_id; |
4cc96131 | 1755 | md->use_blk_mq = dm_use_blk_mq_default(); |
591ddcfc | 1756 | md->init_tio_pdu = false; |
a5664dad | 1757 | md->type = DM_TYPE_NONE; |
e61290a4 | 1758 | mutex_init(&md->suspend_lock); |
a5664dad | 1759 | mutex_init(&md->type_lock); |
86f1152b | 1760 | mutex_init(&md->table_devices_lock); |
022c2611 | 1761 | spin_lock_init(&md->deferred_lock); |
1da177e4 | 1762 | atomic_set(&md->holders, 1); |
5c6bd75d | 1763 | atomic_set(&md->open_count, 0); |
1da177e4 | 1764 | atomic_set(&md->event_nr, 0); |
7a8c3d3b MA |
1765 | atomic_set(&md->uevent_seq, 0); |
1766 | INIT_LIST_HEAD(&md->uevent_list); | |
86f1152b | 1767 | INIT_LIST_HEAD(&md->table_devices); |
7a8c3d3b | 1768 | spin_lock_init(&md->uevent_lock); |
1da177e4 | 1769 | |
115485e8 | 1770 | md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id); |
1da177e4 | 1771 | if (!md->queue) |
0f20972f | 1772 | goto bad; |
1da177e4 | 1773 | |
4a0b4ddf | 1774 | dm_init_md_queue(md); |
9faf400f | 1775 | |
115485e8 | 1776 | md->disk = alloc_disk_node(1, numa_node_id); |
1da177e4 | 1777 | if (!md->disk) |
0f20972f | 1778 | goto bad; |
1da177e4 | 1779 | |
316d315b NK |
1780 | atomic_set(&md->pending[0], 0); |
1781 | atomic_set(&md->pending[1], 0); | |
f0b04115 | 1782 | init_waitqueue_head(&md->wait); |
53d5914f | 1783 | INIT_WORK(&md->work, dm_wq_work); |
f0b04115 | 1784 | init_waitqueue_head(&md->eventq); |
2995fa78 | 1785 | init_completion(&md->kobj_holder.completion); |
2eb6e1e3 | 1786 | md->kworker_task = NULL; |
f0b04115 | 1787 | |
1da177e4 LT |
1788 | md->disk->major = _major; |
1789 | md->disk->first_minor = minor; | |
1790 | md->disk->fops = &dm_blk_dops; | |
1791 | md->disk->queue = md->queue; | |
1792 | md->disk->private_data = md; | |
1793 | sprintf(md->disk->disk_name, "dm-%d", minor); | |
f26c5719 DW |
1794 | |
1795 | dax_dev = alloc_dax(md, md->disk->disk_name, &dm_dax_ops); | |
1796 | if (!dax_dev) | |
1797 | goto bad; | |
1798 | md->dax_dev = dax_dev; | |
1799 | ||
1da177e4 | 1800 | add_disk(md->disk); |
7e51f257 | 1801 | format_dev_t(md->name, MKDEV(_major, minor)); |
1da177e4 | 1802 | |
670368a8 | 1803 | md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0); |
304f3f6a | 1804 | if (!md->wq) |
0f20972f | 1805 | goto bad; |
304f3f6a | 1806 | |
32a926da MP |
1807 | md->bdev = bdget_disk(md->disk, 0); |
1808 | if (!md->bdev) | |
0f20972f | 1809 | goto bad; |
32a926da | 1810 | |
3a83f467 | 1811 | bio_init(&md->flush_bio, NULL, 0); |
74d46992 | 1812 | bio_set_dev(&md->flush_bio, md->bdev); |
ff0361b3 | 1813 | md->flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC; |
6a8736d1 | 1814 | |
fd2ed4d2 MP |
1815 | dm_stats_init(&md->stats); |
1816 | ||
ba61fdd1 | 1817 | /* Populate the mapping, nobody knows we exist yet */ |
f32c10b0 | 1818 | spin_lock(&_minor_lock); |
ba61fdd1 | 1819 | old_md = idr_replace(&_minor_idr, md, minor); |
f32c10b0 | 1820 | spin_unlock(&_minor_lock); |
ba61fdd1 JM |
1821 | |
1822 | BUG_ON(old_md != MINOR_ALLOCED); | |
1823 | ||
1da177e4 LT |
1824 | return md; |
1825 | ||
0f20972f MS |
1826 | bad: |
1827 | cleanup_mapped_device(md); | |
83d5e5b0 | 1828 | bad_io_barrier: |
1da177e4 | 1829 | free_minor(minor); |
6ed7ade8 | 1830 | bad_minor: |
10da4f79 | 1831 | module_put(THIS_MODULE); |
6ed7ade8 | 1832 | bad_module_get: |
856eb091 | 1833 | kvfree(md); |
1da177e4 LT |
1834 | return NULL; |
1835 | } | |
1836 | ||
ae9da83f JN |
1837 | static void unlock_fs(struct mapped_device *md); |
1838 | ||
1da177e4 LT |
1839 | static void free_dev(struct mapped_device *md) |
1840 | { | |
f331c029 | 1841 | int minor = MINOR(disk_devt(md->disk)); |
63d94e48 | 1842 | |
32a926da | 1843 | unlock_fs(md); |
2eb6e1e3 | 1844 | |
0f20972f | 1845 | cleanup_mapped_device(md); |
63a4f065 | 1846 | |
86f1152b | 1847 | free_table_devices(&md->table_devices); |
63a4f065 | 1848 | dm_stats_cleanup(&md->stats); |
63a4f065 MS |
1849 | free_minor(minor); |
1850 | ||
10da4f79 | 1851 | module_put(THIS_MODULE); |
856eb091 | 1852 | kvfree(md); |
1da177e4 LT |
1853 | } |
1854 | ||
e6ee8c0b KU |
1855 | static void __bind_mempools(struct mapped_device *md, struct dm_table *t) |
1856 | { | |
c0820cf5 | 1857 | struct dm_md_mempools *p = dm_table_get_md_mempools(t); |
e6ee8c0b | 1858 | |
0776aa0e | 1859 | if (dm_table_bio_based(t)) { |
64f52b0e MS |
1860 | /* |
1861 | * The md may already have mempools that need changing. | |
1862 | * If so, reload bioset because front_pad may have changed | |
1863 | * because a different table was loaded. | |
1864 | */ | |
0776aa0e | 1865 | if (md->bs) { |
16245bdc | 1866 | bioset_free(md->bs); |
0776aa0e | 1867 | md->bs = NULL; |
16245bdc | 1868 | } |
64f52b0e MS |
1869 | if (md->io_bs) { |
1870 | bioset_free(md->io_bs); | |
1871 | md->io_bs = NULL; | |
1872 | } | |
0776aa0e MS |
1873 | |
1874 | } else if (md->bs) { | |
4e6e36c3 MS |
1875 | /* |
1876 | * There's no need to reload with request-based dm | |
1877 | * because the size of front_pad doesn't change. | |
1878 | * Note for future: If you are to reload bioset, | |
1879 | * prep-ed requests in the queue may refer | |
1880 | * to bio from the old bioset, so you must walk | |
1881 | * through the queue to unprep. | |
1882 | */ | |
1883 | goto out; | |
c0820cf5 | 1884 | } |
e6ee8c0b | 1885 | |
dde1e1ec | 1886 | BUG_ON(!p || md->bs || md->io_bs); |
cbc4e3c1 | 1887 | |
e6ee8c0b KU |
1888 | md->bs = p->bs; |
1889 | p->bs = NULL; | |
64f52b0e MS |
1890 | md->io_bs = p->io_bs; |
1891 | p->io_bs = NULL; | |
e6ee8c0b | 1892 | out: |
02233342 | 1893 | /* mempool bind completed, no longer need any mempools in the table */ |
e6ee8c0b KU |
1894 | dm_table_free_md_mempools(t); |
1895 | } | |
1896 | ||
1da177e4 LT |
1897 | /* |
1898 | * Bind a table to the device. | |
1899 | */ | |
1900 | static void event_callback(void *context) | |
1901 | { | |
7a8c3d3b MA |
1902 | unsigned long flags; |
1903 | LIST_HEAD(uevents); | |
1da177e4 LT |
1904 | struct mapped_device *md = (struct mapped_device *) context; |
1905 | ||
7a8c3d3b MA |
1906 | spin_lock_irqsave(&md->uevent_lock, flags); |
1907 | list_splice_init(&md->uevent_list, &uevents); | |
1908 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
1909 | ||
ed9e1982 | 1910 | dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj); |
7a8c3d3b | 1911 | |
1da177e4 LT |
1912 | atomic_inc(&md->event_nr); |
1913 | wake_up(&md->eventq); | |
62e08243 | 1914 | dm_issue_global_event(); |
1da177e4 LT |
1915 | } |
1916 | ||
c217649b MS |
1917 | /* |
1918 | * Protected by md->suspend_lock obtained by dm_swap_table(). | |
1919 | */ | |
4e90188b | 1920 | static void __set_size(struct mapped_device *md, sector_t size) |
1da177e4 | 1921 | { |
1ea0654e BVA |
1922 | lockdep_assert_held(&md->suspend_lock); |
1923 | ||
4e90188b | 1924 | set_capacity(md->disk, size); |
1da177e4 | 1925 | |
db8fef4f | 1926 | i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT); |
1da177e4 LT |
1927 | } |
1928 | ||
042d2a9b AK |
1929 | /* |
1930 | * Returns old map, which caller must destroy. | |
1931 | */ | |
1932 | static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t, | |
1933 | struct queue_limits *limits) | |
1da177e4 | 1934 | { |
042d2a9b | 1935 | struct dm_table *old_map; |
165125e1 | 1936 | struct request_queue *q = md->queue; |
1da177e4 LT |
1937 | sector_t size; |
1938 | ||
5a8f1f80 BVA |
1939 | lockdep_assert_held(&md->suspend_lock); |
1940 | ||
1da177e4 | 1941 | size = dm_table_get_size(t); |
3ac51e74 DW |
1942 | |
1943 | /* | |
1944 | * Wipe any geometry if the size of the table changed. | |
1945 | */ | |
fd2ed4d2 | 1946 | if (size != dm_get_size(md)) |
3ac51e74 DW |
1947 | memset(&md->geometry, 0, sizeof(md->geometry)); |
1948 | ||
32a926da | 1949 | __set_size(md, size); |
d5816876 | 1950 | |
2ca3310e AK |
1951 | dm_table_event_callback(t, event_callback, md); |
1952 | ||
e6ee8c0b KU |
1953 | /* |
1954 | * The queue hasn't been stopped yet, if the old table type wasn't | |
1955 | * for request-based during suspension. So stop it to prevent | |
1956 | * I/O mapping before resume. | |
1957 | * This must be done before setting the queue restrictions, | |
1958 | * because request-based dm may be run just after the setting. | |
1959 | */ | |
16f12266 | 1960 | if (dm_table_request_based(t)) { |
eca7ee6d | 1961 | dm_stop_queue(q); |
16f12266 MS |
1962 | /* |
1963 | * Leverage the fact that request-based DM targets are | |
1964 | * immutable singletons and establish md->immutable_target | |
1965 | * - used to optimize both dm_request_fn and dm_mq_queue_rq | |
1966 | */ | |
1967 | md->immutable_target = dm_table_get_immutable_target(t); | |
1968 | } | |
e6ee8c0b KU |
1969 | |
1970 | __bind_mempools(md, t); | |
1971 | ||
a12f5d48 | 1972 | old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock)); |
1d3aa6f6 | 1973 | rcu_assign_pointer(md->map, (void *)t); |
36a0456f AK |
1974 | md->immutable_target_type = dm_table_get_immutable_target_type(t); |
1975 | ||
754c5fc7 | 1976 | dm_table_set_restrictions(t, q, limits); |
41abc4e1 HR |
1977 | if (old_map) |
1978 | dm_sync_table(md); | |
1da177e4 | 1979 | |
042d2a9b | 1980 | return old_map; |
1da177e4 LT |
1981 | } |
1982 | ||
a7940155 AK |
1983 | /* |
1984 | * Returns unbound table for the caller to free. | |
1985 | */ | |
1986 | static struct dm_table *__unbind(struct mapped_device *md) | |
1da177e4 | 1987 | { |
a12f5d48 | 1988 | struct dm_table *map = rcu_dereference_protected(md->map, 1); |
1da177e4 LT |
1989 | |
1990 | if (!map) | |
a7940155 | 1991 | return NULL; |
1da177e4 LT |
1992 | |
1993 | dm_table_event_callback(map, NULL, NULL); | |
9cdb8520 | 1994 | RCU_INIT_POINTER(md->map, NULL); |
83d5e5b0 | 1995 | dm_sync_table(md); |
a7940155 AK |
1996 | |
1997 | return map; | |
1da177e4 LT |
1998 | } |
1999 | ||
2000 | /* | |
2001 | * Constructor for a new device. | |
2002 | */ | |
2b06cfff | 2003 | int dm_create(int minor, struct mapped_device **result) |
1da177e4 LT |
2004 | { |
2005 | struct mapped_device *md; | |
2006 | ||
2b06cfff | 2007 | md = alloc_dev(minor); |
1da177e4 LT |
2008 | if (!md) |
2009 | return -ENXIO; | |
2010 | ||
784aae73 MB |
2011 | dm_sysfs_init(md); |
2012 | ||
1da177e4 LT |
2013 | *result = md; |
2014 | return 0; | |
2015 | } | |
2016 | ||
a5664dad MS |
2017 | /* |
2018 | * Functions to manage md->type. | |
2019 | * All are required to hold md->type_lock. | |
2020 | */ | |
2021 | void dm_lock_md_type(struct mapped_device *md) | |
2022 | { | |
2023 | mutex_lock(&md->type_lock); | |
2024 | } | |
2025 | ||
2026 | void dm_unlock_md_type(struct mapped_device *md) | |
2027 | { | |
2028 | mutex_unlock(&md->type_lock); | |
2029 | } | |
2030 | ||
7e0d574f | 2031 | void dm_set_md_type(struct mapped_device *md, enum dm_queue_mode type) |
a5664dad | 2032 | { |
00c4fc3b | 2033 | BUG_ON(!mutex_is_locked(&md->type_lock)); |
a5664dad MS |
2034 | md->type = type; |
2035 | } | |
2036 | ||
7e0d574f | 2037 | enum dm_queue_mode dm_get_md_type(struct mapped_device *md) |
a5664dad MS |
2038 | { |
2039 | return md->type; | |
2040 | } | |
2041 | ||
36a0456f AK |
2042 | struct target_type *dm_get_immutable_target_type(struct mapped_device *md) |
2043 | { | |
2044 | return md->immutable_target_type; | |
2045 | } | |
2046 | ||
f84cb8a4 MS |
2047 | /* |
2048 | * The queue_limits are only valid as long as you have a reference | |
2049 | * count on 'md'. | |
2050 | */ | |
2051 | struct queue_limits *dm_get_queue_limits(struct mapped_device *md) | |
2052 | { | |
2053 | BUG_ON(!atomic_read(&md->holders)); | |
2054 | return &md->queue->limits; | |
2055 | } | |
2056 | EXPORT_SYMBOL_GPL(dm_get_queue_limits); | |
2057 | ||
4a0b4ddf MS |
2058 | /* |
2059 | * Setup the DM device's queue based on md's type | |
2060 | */ | |
591ddcfc | 2061 | int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) |
4a0b4ddf | 2062 | { |
bfebd1cd | 2063 | int r; |
7e0d574f | 2064 | enum dm_queue_mode type = dm_get_md_type(md); |
bfebd1cd | 2065 | |
545ed20e | 2066 | switch (type) { |
bfebd1cd | 2067 | case DM_TYPE_REQUEST_BASED: |
eb8db831 | 2068 | r = dm_old_init_request_queue(md, t); |
bfebd1cd | 2069 | if (r) { |
eca7ee6d | 2070 | DMERR("Cannot initialize queue for request-based mapped device"); |
bfebd1cd | 2071 | return r; |
ff36ab34 | 2072 | } |
bfebd1cd MS |
2073 | break; |
2074 | case DM_TYPE_MQ_REQUEST_BASED: | |
e83068a5 | 2075 | r = dm_mq_init_request_queue(md, t); |
bfebd1cd | 2076 | if (r) { |
eca7ee6d | 2077 | DMERR("Cannot initialize queue for request-based dm-mq mapped device"); |
bfebd1cd MS |
2078 | return r; |
2079 | } | |
2080 | break; | |
2081 | case DM_TYPE_BIO_BASED: | |
545ed20e | 2082 | case DM_TYPE_DAX_BIO_BASED: |
eca7ee6d | 2083 | dm_init_normal_md_queue(md); |
ff36ab34 | 2084 | blk_queue_make_request(md->queue, dm_make_request); |
bfebd1cd | 2085 | break; |
7e0d574f BVA |
2086 | case DM_TYPE_NONE: |
2087 | WARN_ON_ONCE(true); | |
2088 | break; | |
4a0b4ddf MS |
2089 | } |
2090 | ||
2091 | return 0; | |
2092 | } | |
2093 | ||
2bec1f4a | 2094 | struct mapped_device *dm_get_md(dev_t dev) |
1da177e4 LT |
2095 | { |
2096 | struct mapped_device *md; | |
1da177e4 LT |
2097 | unsigned minor = MINOR(dev); |
2098 | ||
2099 | if (MAJOR(dev) != _major || minor >= (1 << MINORBITS)) | |
2100 | return NULL; | |
2101 | ||
f32c10b0 | 2102 | spin_lock(&_minor_lock); |
1da177e4 LT |
2103 | |
2104 | md = idr_find(&_minor_idr, minor); | |
49de5769 MS |
2105 | if (!md || md == MINOR_ALLOCED || (MINOR(disk_devt(dm_disk(md))) != minor) || |
2106 | test_bit(DMF_FREEING, &md->flags) || dm_deleting_md(md)) { | |
2107 | md = NULL; | |
2108 | goto out; | |
fba9f90e | 2109 | } |
49de5769 | 2110 | dm_get(md); |
fba9f90e | 2111 | out: |
f32c10b0 | 2112 | spin_unlock(&_minor_lock); |
1da177e4 | 2113 | |
637842cf DT |
2114 | return md; |
2115 | } | |
3cf2e4ba | 2116 | EXPORT_SYMBOL_GPL(dm_get_md); |
d229a958 | 2117 | |
9ade92a9 | 2118 | void *dm_get_mdptr(struct mapped_device *md) |
637842cf | 2119 | { |
9ade92a9 | 2120 | return md->interface_ptr; |
1da177e4 LT |
2121 | } |
2122 | ||
2123 | void dm_set_mdptr(struct mapped_device *md, void *ptr) | |
2124 | { | |
2125 | md->interface_ptr = ptr; | |
2126 | } | |
2127 | ||
2128 | void dm_get(struct mapped_device *md) | |
2129 | { | |
2130 | atomic_inc(&md->holders); | |
3f77316d | 2131 | BUG_ON(test_bit(DMF_FREEING, &md->flags)); |
1da177e4 LT |
2132 | } |
2133 | ||
09ee96b2 MP |
2134 | int dm_hold(struct mapped_device *md) |
2135 | { | |
2136 | spin_lock(&_minor_lock); | |
2137 | if (test_bit(DMF_FREEING, &md->flags)) { | |
2138 | spin_unlock(&_minor_lock); | |
2139 | return -EBUSY; | |
2140 | } | |
2141 | dm_get(md); | |
2142 | spin_unlock(&_minor_lock); | |
2143 | return 0; | |
2144 | } | |
2145 | EXPORT_SYMBOL_GPL(dm_hold); | |
2146 | ||
72d94861 AK |
2147 | const char *dm_device_name(struct mapped_device *md) |
2148 | { | |
2149 | return md->name; | |
2150 | } | |
2151 | EXPORT_SYMBOL_GPL(dm_device_name); | |
2152 | ||
3f77316d | 2153 | static void __dm_destroy(struct mapped_device *md, bool wait) |
1da177e4 | 2154 | { |
3b785fbc | 2155 | struct request_queue *q = dm_get_md_queue(md); |
1134e5ae | 2156 | struct dm_table *map; |
83d5e5b0 | 2157 | int srcu_idx; |
1da177e4 | 2158 | |
3f77316d | 2159 | might_sleep(); |
fba9f90e | 2160 | |
63a4f065 | 2161 | spin_lock(&_minor_lock); |
3f77316d KU |
2162 | idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); |
2163 | set_bit(DMF_FREEING, &md->flags); | |
2164 | spin_unlock(&_minor_lock); | |
3b785fbc | 2165 | |
2e91c369 | 2166 | blk_set_queue_dying(q); |
3f77316d | 2167 | |
02233342 | 2168 | if (dm_request_based(md) && md->kworker_task) |
3989144f | 2169 | kthread_flush_worker(&md->kworker); |
2eb6e1e3 | 2170 | |
ab7c7bb6 MP |
2171 | /* |
2172 | * Take suspend_lock so that presuspend and postsuspend methods | |
2173 | * do not race with internal suspend. | |
2174 | */ | |
2175 | mutex_lock(&md->suspend_lock); | |
2a708cff | 2176 | map = dm_get_live_table(md, &srcu_idx); |
3f77316d KU |
2177 | if (!dm_suspended_md(md)) { |
2178 | dm_table_presuspend_targets(map); | |
2179 | dm_table_postsuspend_targets(map); | |
1da177e4 | 2180 | } |
83d5e5b0 MP |
2181 | /* dm_put_live_table must be before msleep, otherwise deadlock is possible */ |
2182 | dm_put_live_table(md, srcu_idx); | |
2a708cff | 2183 | mutex_unlock(&md->suspend_lock); |
83d5e5b0 | 2184 | |
3f77316d KU |
2185 | /* |
2186 | * Rare, but there may be I/O requests still going to complete, | |
2187 | * for example. Wait for all references to disappear. | |
2188 | * No one should increment the reference count of the mapped_device, | |
2189 | * after the mapped_device state becomes DMF_FREEING. | |
2190 | */ | |
2191 | if (wait) | |
2192 | while (atomic_read(&md->holders)) | |
2193 | msleep(1); | |
2194 | else if (atomic_read(&md->holders)) | |
2195 | DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)", | |
2196 | dm_device_name(md), atomic_read(&md->holders)); | |
2197 | ||
2198 | dm_sysfs_exit(md); | |
3f77316d KU |
2199 | dm_table_destroy(__unbind(md)); |
2200 | free_dev(md); | |
2201 | } | |
2202 | ||
2203 | void dm_destroy(struct mapped_device *md) | |
2204 | { | |
2205 | __dm_destroy(md, true); | |
2206 | } | |
2207 | ||
2208 | void dm_destroy_immediate(struct mapped_device *md) | |
2209 | { | |
2210 | __dm_destroy(md, false); | |
2211 | } | |
2212 | ||
2213 | void dm_put(struct mapped_device *md) | |
2214 | { | |
2215 | atomic_dec(&md->holders); | |
1da177e4 | 2216 | } |
79eb885c | 2217 | EXPORT_SYMBOL_GPL(dm_put); |
1da177e4 | 2218 | |
b48633f8 | 2219 | static int dm_wait_for_completion(struct mapped_device *md, long task_state) |
46125c1c MB |
2220 | { |
2221 | int r = 0; | |
9f4c3f87 | 2222 | DEFINE_WAIT(wait); |
46125c1c MB |
2223 | |
2224 | while (1) { | |
9f4c3f87 | 2225 | prepare_to_wait(&md->wait, &wait, task_state); |
46125c1c | 2226 | |
b4324fee | 2227 | if (!md_in_flight(md)) |
46125c1c MB |
2228 | break; |
2229 | ||
e3fabdfd | 2230 | if (signal_pending_state(task_state, current)) { |
46125c1c MB |
2231 | r = -EINTR; |
2232 | break; | |
2233 | } | |
2234 | ||
2235 | io_schedule(); | |
2236 | } | |
9f4c3f87 | 2237 | finish_wait(&md->wait, &wait); |
b44ebeb0 | 2238 | |
46125c1c MB |
2239 | return r; |
2240 | } | |
2241 | ||
1da177e4 LT |
2242 | /* |
2243 | * Process the deferred bios | |
2244 | */ | |
ef208587 | 2245 | static void dm_wq_work(struct work_struct *work) |
1da177e4 | 2246 | { |
ef208587 MP |
2247 | struct mapped_device *md = container_of(work, struct mapped_device, |
2248 | work); | |
6d6f10df | 2249 | struct bio *c; |
83d5e5b0 MP |
2250 | int srcu_idx; |
2251 | struct dm_table *map; | |
1da177e4 | 2252 | |
83d5e5b0 | 2253 | map = dm_get_live_table(md, &srcu_idx); |
ef208587 | 2254 | |
3b00b203 | 2255 | while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) { |
df12ee99 AK |
2256 | spin_lock_irq(&md->deferred_lock); |
2257 | c = bio_list_pop(&md->deferred); | |
2258 | spin_unlock_irq(&md->deferred_lock); | |
2259 | ||
6a8736d1 | 2260 | if (!c) |
df12ee99 | 2261 | break; |
022c2611 | 2262 | |
e6ee8c0b KU |
2263 | if (dm_request_based(md)) |
2264 | generic_make_request(c); | |
6a8736d1 | 2265 | else |
83d5e5b0 | 2266 | __split_and_process_bio(md, map, c); |
022c2611 | 2267 | } |
73d410c0 | 2268 | |
83d5e5b0 | 2269 | dm_put_live_table(md, srcu_idx); |
1da177e4 LT |
2270 | } |
2271 | ||
9a1fb464 | 2272 | static void dm_queue_flush(struct mapped_device *md) |
304f3f6a | 2273 | { |
3b00b203 | 2274 | clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
4e857c58 | 2275 | smp_mb__after_atomic(); |
53d5914f | 2276 | queue_work(md->wq, &md->work); |
304f3f6a MB |
2277 | } |
2278 | ||
1da177e4 | 2279 | /* |
042d2a9b | 2280 | * Swap in a new table, returning the old one for the caller to destroy. |
1da177e4 | 2281 | */ |
042d2a9b | 2282 | struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table) |
1da177e4 | 2283 | { |
87eb5b21 | 2284 | struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL); |
754c5fc7 | 2285 | struct queue_limits limits; |
042d2a9b | 2286 | int r; |
1da177e4 | 2287 | |
e61290a4 | 2288 | mutex_lock(&md->suspend_lock); |
1da177e4 LT |
2289 | |
2290 | /* device must be suspended */ | |
4f186f8b | 2291 | if (!dm_suspended_md(md)) |
93c534ae | 2292 | goto out; |
1da177e4 | 2293 | |
3ae70656 MS |
2294 | /* |
2295 | * If the new table has no data devices, retain the existing limits. | |
2296 | * This helps multipath with queue_if_no_path if all paths disappear, | |
2297 | * then new I/O is queued based on these limits, and then some paths | |
2298 | * reappear. | |
2299 | */ | |
2300 | if (dm_table_has_no_data_devices(table)) { | |
83d5e5b0 | 2301 | live_map = dm_get_live_table_fast(md); |
3ae70656 MS |
2302 | if (live_map) |
2303 | limits = md->queue->limits; | |
83d5e5b0 | 2304 | dm_put_live_table_fast(md); |
3ae70656 MS |
2305 | } |
2306 | ||
87eb5b21 MC |
2307 | if (!live_map) { |
2308 | r = dm_calculate_queue_limits(table, &limits); | |
2309 | if (r) { | |
2310 | map = ERR_PTR(r); | |
2311 | goto out; | |
2312 | } | |
042d2a9b | 2313 | } |
754c5fc7 | 2314 | |
042d2a9b | 2315 | map = __bind(md, table, &limits); |
62e08243 | 2316 | dm_issue_global_event(); |
1da177e4 | 2317 | |
93c534ae | 2318 | out: |
e61290a4 | 2319 | mutex_unlock(&md->suspend_lock); |
042d2a9b | 2320 | return map; |
1da177e4 LT |
2321 | } |
2322 | ||
2323 | /* | |
2324 | * Functions to lock and unlock any filesystem running on the | |
2325 | * device. | |
2326 | */ | |
2ca3310e | 2327 | static int lock_fs(struct mapped_device *md) |
1da177e4 | 2328 | { |
e39e2e95 | 2329 | int r; |
1da177e4 LT |
2330 | |
2331 | WARN_ON(md->frozen_sb); | |
dfbe03f6 | 2332 | |
db8fef4f | 2333 | md->frozen_sb = freeze_bdev(md->bdev); |
dfbe03f6 | 2334 | if (IS_ERR(md->frozen_sb)) { |
cf222b37 | 2335 | r = PTR_ERR(md->frozen_sb); |
e39e2e95 AK |
2336 | md->frozen_sb = NULL; |
2337 | return r; | |
dfbe03f6 AK |
2338 | } |
2339 | ||
aa8d7c2f AK |
2340 | set_bit(DMF_FROZEN, &md->flags); |
2341 | ||
1da177e4 LT |
2342 | return 0; |
2343 | } | |
2344 | ||
2ca3310e | 2345 | static void unlock_fs(struct mapped_device *md) |
1da177e4 | 2346 | { |
aa8d7c2f AK |
2347 | if (!test_bit(DMF_FROZEN, &md->flags)) |
2348 | return; | |
2349 | ||
db8fef4f | 2350 | thaw_bdev(md->bdev, md->frozen_sb); |
1da177e4 | 2351 | md->frozen_sb = NULL; |
aa8d7c2f | 2352 | clear_bit(DMF_FROZEN, &md->flags); |
1da177e4 LT |
2353 | } |
2354 | ||
2355 | /* | |
b48633f8 BVA |
2356 | * @suspend_flags: DM_SUSPEND_LOCKFS_FLAG and/or DM_SUSPEND_NOFLUSH_FLAG |
2357 | * @task_state: e.g. TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE | |
2358 | * @dmf_suspended_flag: DMF_SUSPENDED or DMF_SUSPENDED_INTERNALLY | |
2359 | * | |
ffcc3936 MS |
2360 | * If __dm_suspend returns 0, the device is completely quiescent |
2361 | * now. There is no request-processing activity. All new requests | |
2362 | * are being added to md->deferred list. | |
cec47e3d | 2363 | */ |
ffcc3936 | 2364 | static int __dm_suspend(struct mapped_device *md, struct dm_table *map, |
b48633f8 | 2365 | unsigned suspend_flags, long task_state, |
eaf9a736 | 2366 | int dmf_suspended_flag) |
1da177e4 | 2367 | { |
ffcc3936 MS |
2368 | bool do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG; |
2369 | bool noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG; | |
2370 | int r; | |
1da177e4 | 2371 | |
5a8f1f80 BVA |
2372 | lockdep_assert_held(&md->suspend_lock); |
2373 | ||
2e93ccc1 KU |
2374 | /* |
2375 | * DMF_NOFLUSH_SUSPENDING must be set before presuspend. | |
2376 | * This flag is cleared before dm_suspend returns. | |
2377 | */ | |
2378 | if (noflush) | |
2379 | set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); | |
86331f39 BVA |
2380 | else |
2381 | pr_debug("%s: suspending with flush\n", dm_device_name(md)); | |
2e93ccc1 | 2382 | |
d67ee213 MS |
2383 | /* |
2384 | * This gets reverted if there's an error later and the targets | |
2385 | * provide the .presuspend_undo hook. | |
2386 | */ | |
cf222b37 AK |
2387 | dm_table_presuspend_targets(map); |
2388 | ||
32a926da | 2389 | /* |
9f518b27 KU |
2390 | * Flush I/O to the device. |
2391 | * Any I/O submitted after lock_fs() may not be flushed. | |
2392 | * noflush takes precedence over do_lockfs. | |
2393 | * (lock_fs() flushes I/Os and waits for them to complete.) | |
32a926da MP |
2394 | */ |
2395 | if (!noflush && do_lockfs) { | |
2396 | r = lock_fs(md); | |
d67ee213 MS |
2397 | if (r) { |
2398 | dm_table_presuspend_undo_targets(map); | |
ffcc3936 | 2399 | return r; |
d67ee213 | 2400 | } |
aa8d7c2f | 2401 | } |
1da177e4 LT |
2402 | |
2403 | /* | |
3b00b203 MP |
2404 | * Here we must make sure that no processes are submitting requests |
2405 | * to target drivers i.e. no one may be executing | |
2406 | * __split_and_process_bio. This is called from dm_request and | |
2407 | * dm_wq_work. | |
2408 | * | |
2409 | * To get all processes out of __split_and_process_bio in dm_request, | |
2410 | * we take the write lock. To prevent any process from reentering | |
6a8736d1 TH |
2411 | * __split_and_process_bio from dm_request and quiesce the thread |
2412 | * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call | |
2413 | * flush_workqueue(md->wq). | |
1da177e4 | 2414 | */ |
1eb787ec | 2415 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); |
41abc4e1 HR |
2416 | if (map) |
2417 | synchronize_srcu(&md->io_barrier); | |
1da177e4 | 2418 | |
d0bcb878 | 2419 | /* |
29e4013d TH |
2420 | * Stop md->queue before flushing md->wq in case request-based |
2421 | * dm defers requests to md->wq from md->queue. | |
d0bcb878 | 2422 | */ |
2eb6e1e3 | 2423 | if (dm_request_based(md)) { |
eca7ee6d | 2424 | dm_stop_queue(md->queue); |
02233342 | 2425 | if (md->kworker_task) |
3989144f | 2426 | kthread_flush_worker(&md->kworker); |
2eb6e1e3 | 2427 | } |
cec47e3d | 2428 | |
d0bcb878 KU |
2429 | flush_workqueue(md->wq); |
2430 | ||
1da177e4 | 2431 | /* |
3b00b203 MP |
2432 | * At this point no more requests are entering target request routines. |
2433 | * We call dm_wait_for_completion to wait for all existing requests | |
2434 | * to finish. | |
1da177e4 | 2435 | */ |
b48633f8 | 2436 | r = dm_wait_for_completion(md, task_state); |
eaf9a736 MS |
2437 | if (!r) |
2438 | set_bit(dmf_suspended_flag, &md->flags); | |
1da177e4 | 2439 | |
6d6f10df | 2440 | if (noflush) |
022c2611 | 2441 | clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags); |
41abc4e1 HR |
2442 | if (map) |
2443 | synchronize_srcu(&md->io_barrier); | |
2e93ccc1 | 2444 | |
1da177e4 | 2445 | /* were we interrupted ? */ |
46125c1c | 2446 | if (r < 0) { |
9a1fb464 | 2447 | dm_queue_flush(md); |
73d410c0 | 2448 | |
cec47e3d | 2449 | if (dm_request_based(md)) |
eca7ee6d | 2450 | dm_start_queue(md->queue); |
cec47e3d | 2451 | |
2ca3310e | 2452 | unlock_fs(md); |
d67ee213 | 2453 | dm_table_presuspend_undo_targets(map); |
ffcc3936 | 2454 | /* pushback list is already flushed, so skip flush */ |
2ca3310e | 2455 | } |
1da177e4 | 2456 | |
ffcc3936 MS |
2457 | return r; |
2458 | } | |
2459 | ||
2460 | /* | |
2461 | * We need to be able to change a mapping table under a mounted | |
2462 | * filesystem. For example we might want to move some data in | |
2463 | * the background. Before the table can be swapped with | |
2464 | * dm_bind_table, dm_suspend must be called to flush any in | |
2465 | * flight bios and ensure that any further io gets deferred. | |
2466 | */ | |
2467 | /* | |
2468 | * Suspend mechanism in request-based dm. | |
2469 | * | |
2470 | * 1. Flush all I/Os by lock_fs() if needed. | |
2471 | * 2. Stop dispatching any I/O by stopping the request_queue. | |
2472 | * 3. Wait for all in-flight I/Os to be completed or requeued. | |
2473 | * | |
2474 | * To abort suspend, start the request_queue. | |
2475 | */ | |
2476 | int dm_suspend(struct mapped_device *md, unsigned suspend_flags) | |
2477 | { | |
2478 | struct dm_table *map = NULL; | |
2479 | int r = 0; | |
2480 | ||
2481 | retry: | |
2482 | mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING); | |
2483 | ||
2484 | if (dm_suspended_md(md)) { | |
2485 | r = -EINVAL; | |
2486 | goto out_unlock; | |
2487 | } | |
2488 | ||
2489 | if (dm_suspended_internally_md(md)) { | |
2490 | /* already internally suspended, wait for internal resume */ | |
2491 | mutex_unlock(&md->suspend_lock); | |
2492 | r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE); | |
2493 | if (r) | |
2494 | return r; | |
2495 | goto retry; | |
2496 | } | |
2497 | ||
a12f5d48 | 2498 | map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock)); |
ffcc3936 | 2499 | |
eaf9a736 | 2500 | r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED); |
ffcc3936 MS |
2501 | if (r) |
2502 | goto out_unlock; | |
3b00b203 | 2503 | |
4d4471cb KU |
2504 | dm_table_postsuspend_targets(map); |
2505 | ||
d287483d | 2506 | out_unlock: |
e61290a4 | 2507 | mutex_unlock(&md->suspend_lock); |
cf222b37 | 2508 | return r; |
1da177e4 LT |
2509 | } |
2510 | ||
ffcc3936 MS |
2511 | static int __dm_resume(struct mapped_device *md, struct dm_table *map) |
2512 | { | |
2513 | if (map) { | |
2514 | int r = dm_table_resume_targets(map); | |
2515 | if (r) | |
2516 | return r; | |
2517 | } | |
2518 | ||
2519 | dm_queue_flush(md); | |
2520 | ||
2521 | /* | |
2522 | * Flushing deferred I/Os must be done after targets are resumed | |
2523 | * so that mapping of targets can work correctly. | |
2524 | * Request-based dm is queueing the deferred I/Os in its request_queue. | |
2525 | */ | |
2526 | if (dm_request_based(md)) | |
eca7ee6d | 2527 | dm_start_queue(md->queue); |
ffcc3936 MS |
2528 | |
2529 | unlock_fs(md); | |
2530 | ||
2531 | return 0; | |
2532 | } | |
2533 | ||
1da177e4 LT |
2534 | int dm_resume(struct mapped_device *md) |
2535 | { | |
8dc23658 | 2536 | int r; |
cf222b37 | 2537 | struct dm_table *map = NULL; |
1da177e4 | 2538 | |
ffcc3936 | 2539 | retry: |
8dc23658 | 2540 | r = -EINVAL; |
ffcc3936 MS |
2541 | mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING); |
2542 | ||
4f186f8b | 2543 | if (!dm_suspended_md(md)) |
cf222b37 | 2544 | goto out; |
cf222b37 | 2545 | |
ffcc3936 MS |
2546 | if (dm_suspended_internally_md(md)) { |
2547 | /* already internally suspended, wait for internal resume */ | |
2548 | mutex_unlock(&md->suspend_lock); | |
2549 | r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE); | |
2550 | if (r) | |
2551 | return r; | |
2552 | goto retry; | |
2553 | } | |
2554 | ||
a12f5d48 | 2555 | map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock)); |
2ca3310e | 2556 | if (!map || !dm_table_get_size(map)) |
cf222b37 | 2557 | goto out; |
1da177e4 | 2558 | |
ffcc3936 | 2559 | r = __dm_resume(md, map); |
8757b776 MB |
2560 | if (r) |
2561 | goto out; | |
2ca3310e | 2562 | |
2ca3310e | 2563 | clear_bit(DMF_SUSPENDED, &md->flags); |
cf222b37 | 2564 | out: |
e61290a4 | 2565 | mutex_unlock(&md->suspend_lock); |
2ca3310e | 2566 | |
cf222b37 | 2567 | return r; |
1da177e4 LT |
2568 | } |
2569 | ||
fd2ed4d2 MP |
2570 | /* |
2571 | * Internal suspend/resume works like userspace-driven suspend. It waits | |
2572 | * until all bios finish and prevents issuing new bios to the target drivers. | |
2573 | * It may be used only from the kernel. | |
fd2ed4d2 MP |
2574 | */ |
2575 | ||
ffcc3936 | 2576 | static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags) |
fd2ed4d2 | 2577 | { |
ffcc3936 MS |
2578 | struct dm_table *map = NULL; |
2579 | ||
1ea0654e BVA |
2580 | lockdep_assert_held(&md->suspend_lock); |
2581 | ||
96b26c8c | 2582 | if (md->internal_suspend_count++) |
ffcc3936 MS |
2583 | return; /* nested internal suspend */ |
2584 | ||
2585 | if (dm_suspended_md(md)) { | |
2586 | set_bit(DMF_SUSPENDED_INTERNALLY, &md->flags); | |
2587 | return; /* nest suspend */ | |
2588 | } | |
2589 | ||
a12f5d48 | 2590 | map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock)); |
ffcc3936 MS |
2591 | |
2592 | /* | |
2593 | * Using TASK_UNINTERRUPTIBLE because only NOFLUSH internal suspend is | |
2594 | * supported. Properly supporting a TASK_INTERRUPTIBLE internal suspend | |
2595 | * would require changing .presuspend to return an error -- avoid this | |
2596 | * until there is a need for more elaborate variants of internal suspend. | |
2597 | */ | |
eaf9a736 MS |
2598 | (void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE, |
2599 | DMF_SUSPENDED_INTERNALLY); | |
ffcc3936 MS |
2600 | |
2601 | dm_table_postsuspend_targets(map); | |
2602 | } | |
2603 | ||
2604 | static void __dm_internal_resume(struct mapped_device *md) | |
2605 | { | |
96b26c8c MP |
2606 | BUG_ON(!md->internal_suspend_count); |
2607 | ||
2608 | if (--md->internal_suspend_count) | |
ffcc3936 MS |
2609 | return; /* resume from nested internal suspend */ |
2610 | ||
fd2ed4d2 | 2611 | if (dm_suspended_md(md)) |
ffcc3936 MS |
2612 | goto done; /* resume from nested suspend */ |
2613 | ||
2614 | /* | |
2615 | * NOTE: existing callers don't need to call dm_table_resume_targets | |
2616 | * (which may fail -- so best to avoid it for now by passing NULL map) | |
2617 | */ | |
2618 | (void) __dm_resume(md, NULL); | |
2619 | ||
2620 | done: | |
2621 | clear_bit(DMF_SUSPENDED_INTERNALLY, &md->flags); | |
2622 | smp_mb__after_atomic(); | |
2623 | wake_up_bit(&md->flags, DMF_SUSPENDED_INTERNALLY); | |
2624 | } | |
2625 | ||
2626 | void dm_internal_suspend_noflush(struct mapped_device *md) | |
2627 | { | |
2628 | mutex_lock(&md->suspend_lock); | |
2629 | __dm_internal_suspend(md, DM_SUSPEND_NOFLUSH_FLAG); | |
2630 | mutex_unlock(&md->suspend_lock); | |
2631 | } | |
2632 | EXPORT_SYMBOL_GPL(dm_internal_suspend_noflush); | |
2633 | ||
2634 | void dm_internal_resume(struct mapped_device *md) | |
2635 | { | |
2636 | mutex_lock(&md->suspend_lock); | |
2637 | __dm_internal_resume(md); | |
2638 | mutex_unlock(&md->suspend_lock); | |
2639 | } | |
2640 | EXPORT_SYMBOL_GPL(dm_internal_resume); | |
2641 | ||
2642 | /* | |
2643 | * Fast variants of internal suspend/resume hold md->suspend_lock, | |
2644 | * which prevents interaction with userspace-driven suspend. | |
2645 | */ | |
2646 | ||
2647 | void dm_internal_suspend_fast(struct mapped_device *md) | |
2648 | { | |
2649 | mutex_lock(&md->suspend_lock); | |
2650 | if (dm_suspended_md(md) || dm_suspended_internally_md(md)) | |
fd2ed4d2 MP |
2651 | return; |
2652 | ||
2653 | set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags); | |
2654 | synchronize_srcu(&md->io_barrier); | |
2655 | flush_workqueue(md->wq); | |
2656 | dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE); | |
2657 | } | |
b735fede | 2658 | EXPORT_SYMBOL_GPL(dm_internal_suspend_fast); |
fd2ed4d2 | 2659 | |
ffcc3936 | 2660 | void dm_internal_resume_fast(struct mapped_device *md) |
fd2ed4d2 | 2661 | { |
ffcc3936 | 2662 | if (dm_suspended_md(md) || dm_suspended_internally_md(md)) |
fd2ed4d2 MP |
2663 | goto done; |
2664 | ||
2665 | dm_queue_flush(md); | |
2666 | ||
2667 | done: | |
2668 | mutex_unlock(&md->suspend_lock); | |
2669 | } | |
b735fede | 2670 | EXPORT_SYMBOL_GPL(dm_internal_resume_fast); |
fd2ed4d2 | 2671 | |
1da177e4 LT |
2672 | /*----------------------------------------------------------------- |
2673 | * Event notification. | |
2674 | *---------------------------------------------------------------*/ | |
3abf85b5 | 2675 | int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action, |
60935eb2 | 2676 | unsigned cookie) |
69267a30 | 2677 | { |
60935eb2 MB |
2678 | char udev_cookie[DM_COOKIE_LENGTH]; |
2679 | char *envp[] = { udev_cookie, NULL }; | |
2680 | ||
2681 | if (!cookie) | |
3abf85b5 | 2682 | return kobject_uevent(&disk_to_dev(md->disk)->kobj, action); |
60935eb2 MB |
2683 | else { |
2684 | snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u", | |
2685 | DM_COOKIE_ENV_VAR_NAME, cookie); | |
3abf85b5 PR |
2686 | return kobject_uevent_env(&disk_to_dev(md->disk)->kobj, |
2687 | action, envp); | |
60935eb2 | 2688 | } |
69267a30 AK |
2689 | } |
2690 | ||
7a8c3d3b MA |
2691 | uint32_t dm_next_uevent_seq(struct mapped_device *md) |
2692 | { | |
2693 | return atomic_add_return(1, &md->uevent_seq); | |
2694 | } | |
2695 | ||
1da177e4 LT |
2696 | uint32_t dm_get_event_nr(struct mapped_device *md) |
2697 | { | |
2698 | return atomic_read(&md->event_nr); | |
2699 | } | |
2700 | ||
2701 | int dm_wait_event(struct mapped_device *md, int event_nr) | |
2702 | { | |
2703 | return wait_event_interruptible(md->eventq, | |
2704 | (event_nr != atomic_read(&md->event_nr))); | |
2705 | } | |
2706 | ||
7a8c3d3b MA |
2707 | void dm_uevent_add(struct mapped_device *md, struct list_head *elist) |
2708 | { | |
2709 | unsigned long flags; | |
2710 | ||
2711 | spin_lock_irqsave(&md->uevent_lock, flags); | |
2712 | list_add(elist, &md->uevent_list); | |
2713 | spin_unlock_irqrestore(&md->uevent_lock, flags); | |
2714 | } | |
2715 | ||
1da177e4 LT |
2716 | /* |
2717 | * The gendisk is only valid as long as you have a reference | |
2718 | * count on 'md'. | |
2719 | */ | |
2720 | struct gendisk *dm_disk(struct mapped_device *md) | |
2721 | { | |
2722 | return md->disk; | |
2723 | } | |
65ff5b7d | 2724 | EXPORT_SYMBOL_GPL(dm_disk); |
1da177e4 | 2725 | |
784aae73 MB |
2726 | struct kobject *dm_kobject(struct mapped_device *md) |
2727 | { | |
2995fa78 | 2728 | return &md->kobj_holder.kobj; |
784aae73 MB |
2729 | } |
2730 | ||
784aae73 MB |
2731 | struct mapped_device *dm_get_from_kobject(struct kobject *kobj) |
2732 | { | |
2733 | struct mapped_device *md; | |
2734 | ||
2995fa78 | 2735 | md = container_of(kobj, struct mapped_device, kobj_holder.kobj); |
784aae73 | 2736 | |
b9a41d21 HT |
2737 | spin_lock(&_minor_lock); |
2738 | if (test_bit(DMF_FREEING, &md->flags) || dm_deleting_md(md)) { | |
2739 | md = NULL; | |
2740 | goto out; | |
2741 | } | |
784aae73 | 2742 | dm_get(md); |
b9a41d21 HT |
2743 | out: |
2744 | spin_unlock(&_minor_lock); | |
2745 | ||
784aae73 MB |
2746 | return md; |
2747 | } | |
2748 | ||
4f186f8b | 2749 | int dm_suspended_md(struct mapped_device *md) |
1da177e4 LT |
2750 | { |
2751 | return test_bit(DMF_SUSPENDED, &md->flags); | |
2752 | } | |
2753 | ||
ffcc3936 MS |
2754 | int dm_suspended_internally_md(struct mapped_device *md) |
2755 | { | |
2756 | return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags); | |
2757 | } | |
2758 | ||
2c140a24 MP |
2759 | int dm_test_deferred_remove_flag(struct mapped_device *md) |
2760 | { | |
2761 | return test_bit(DMF_DEFERRED_REMOVE, &md->flags); | |
2762 | } | |
2763 | ||
64dbce58 KU |
2764 | int dm_suspended(struct dm_target *ti) |
2765 | { | |
ecdb2e25 | 2766 | return dm_suspended_md(dm_table_get_md(ti->table)); |
64dbce58 KU |
2767 | } |
2768 | EXPORT_SYMBOL_GPL(dm_suspended); | |
2769 | ||
2e93ccc1 KU |
2770 | int dm_noflush_suspending(struct dm_target *ti) |
2771 | { | |
ecdb2e25 | 2772 | return __noflush_suspending(dm_table_get_md(ti->table)); |
2e93ccc1 KU |
2773 | } |
2774 | EXPORT_SYMBOL_GPL(dm_noflush_suspending); | |
2775 | ||
7e0d574f | 2776 | struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type, |
0776aa0e MS |
2777 | unsigned integrity, unsigned per_io_data_size, |
2778 | unsigned min_pool_size) | |
e6ee8c0b | 2779 | { |
115485e8 | 2780 | struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id); |
78d8e58a | 2781 | unsigned int pool_size = 0; |
64f52b0e | 2782 | unsigned int front_pad, io_front_pad; |
e6ee8c0b KU |
2783 | |
2784 | if (!pools) | |
4e6e36c3 | 2785 | return NULL; |
e6ee8c0b | 2786 | |
78d8e58a MS |
2787 | switch (type) { |
2788 | case DM_TYPE_BIO_BASED: | |
545ed20e | 2789 | case DM_TYPE_DAX_BIO_BASED: |
0776aa0e | 2790 | pool_size = max(dm_get_reserved_bio_based_ios(), min_pool_size); |
30187e1d | 2791 | front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone); |
64f52b0e MS |
2792 | io_front_pad = roundup(front_pad, __alignof__(struct dm_io)) + offsetof(struct dm_io, tio); |
2793 | pools->io_bs = bioset_create(pool_size, io_front_pad, 0); | |
2794 | if (!pools->io_bs) | |
2795 | goto out; | |
2796 | if (integrity && bioset_integrity_create(pools->io_bs, pool_size)) | |
2797 | goto out; | |
78d8e58a MS |
2798 | break; |
2799 | case DM_TYPE_REQUEST_BASED: | |
78d8e58a | 2800 | case DM_TYPE_MQ_REQUEST_BASED: |
0776aa0e | 2801 | pool_size = max(dm_get_reserved_rq_based_ios(), min_pool_size); |
78d8e58a | 2802 | front_pad = offsetof(struct dm_rq_clone_bio_info, clone); |
591ddcfc | 2803 | /* per_io_data_size is used for blk-mq pdu at queue allocation */ |
78d8e58a MS |
2804 | break; |
2805 | default: | |
2806 | BUG(); | |
2807 | } | |
2808 | ||
4a3f54d9 | 2809 | pools->bs = bioset_create(pool_size, front_pad, 0); |
e6ee8c0b | 2810 | if (!pools->bs) |
5f015204 | 2811 | goto out; |
e6ee8c0b | 2812 | |
a91a2785 | 2813 | if (integrity && bioset_integrity_create(pools->bs, pool_size)) |
5f015204 | 2814 | goto out; |
a91a2785 | 2815 | |
e6ee8c0b | 2816 | return pools; |
5f1b670d | 2817 | |
5f1b670d CH |
2818 | out: |
2819 | dm_free_md_mempools(pools); | |
78d8e58a | 2820 | |
4e6e36c3 | 2821 | return NULL; |
e6ee8c0b KU |
2822 | } |
2823 | ||
2824 | void dm_free_md_mempools(struct dm_md_mempools *pools) | |
2825 | { | |
2826 | if (!pools) | |
2827 | return; | |
2828 | ||
e6ee8c0b KU |
2829 | if (pools->bs) |
2830 | bioset_free(pools->bs); | |
64f52b0e MS |
2831 | if (pools->io_bs) |
2832 | bioset_free(pools->io_bs); | |
e6ee8c0b KU |
2833 | |
2834 | kfree(pools); | |
2835 | } | |
2836 | ||
9c72bad1 CH |
2837 | struct dm_pr { |
2838 | u64 old_key; | |
2839 | u64 new_key; | |
2840 | u32 flags; | |
2841 | bool fail_early; | |
2842 | }; | |
2843 | ||
2844 | static int dm_call_pr(struct block_device *bdev, iterate_devices_callout_fn fn, | |
2845 | void *data) | |
71cdb697 CH |
2846 | { |
2847 | struct mapped_device *md = bdev->bd_disk->private_data; | |
9c72bad1 CH |
2848 | struct dm_table *table; |
2849 | struct dm_target *ti; | |
2850 | int ret = -ENOTTY, srcu_idx; | |
71cdb697 | 2851 | |
9c72bad1 CH |
2852 | table = dm_get_live_table(md, &srcu_idx); |
2853 | if (!table || !dm_table_get_size(table)) | |
2854 | goto out; | |
71cdb697 | 2855 | |
9c72bad1 CH |
2856 | /* We only support devices that have a single target */ |
2857 | if (dm_table_get_num_targets(table) != 1) | |
2858 | goto out; | |
2859 | ti = dm_table_get_target(table, 0); | |
71cdb697 | 2860 | |
9c72bad1 CH |
2861 | ret = -EINVAL; |
2862 | if (!ti->type->iterate_devices) | |
2863 | goto out; | |
2864 | ||
2865 | ret = ti->type->iterate_devices(ti, fn, data); | |
2866 | out: | |
2867 | dm_put_live_table(md, srcu_idx); | |
2868 | return ret; | |
2869 | } | |
2870 | ||
2871 | /* | |
2872 | * For register / unregister we need to manually call out to every path. | |
2873 | */ | |
2874 | static int __dm_pr_register(struct dm_target *ti, struct dm_dev *dev, | |
2875 | sector_t start, sector_t len, void *data) | |
2876 | { | |
2877 | struct dm_pr *pr = data; | |
2878 | const struct pr_ops *ops = dev->bdev->bd_disk->fops->pr_ops; | |
2879 | ||
2880 | if (!ops || !ops->pr_register) | |
2881 | return -EOPNOTSUPP; | |
2882 | return ops->pr_register(dev->bdev, pr->old_key, pr->new_key, pr->flags); | |
2883 | } | |
2884 | ||
2885 | static int dm_pr_register(struct block_device *bdev, u64 old_key, u64 new_key, | |
2886 | u32 flags) | |
2887 | { | |
2888 | struct dm_pr pr = { | |
2889 | .old_key = old_key, | |
2890 | .new_key = new_key, | |
2891 | .flags = flags, | |
2892 | .fail_early = true, | |
2893 | }; | |
2894 | int ret; | |
2895 | ||
2896 | ret = dm_call_pr(bdev, __dm_pr_register, &pr); | |
2897 | if (ret && new_key) { | |
2898 | /* unregister all paths if we failed to register any path */ | |
2899 | pr.old_key = new_key; | |
2900 | pr.new_key = 0; | |
2901 | pr.flags = 0; | |
2902 | pr.fail_early = false; | |
2903 | dm_call_pr(bdev, __dm_pr_register, &pr); | |
2904 | } | |
2905 | ||
2906 | return ret; | |
71cdb697 CH |
2907 | } |
2908 | ||
2909 | static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type, | |
956a4025 | 2910 | u32 flags) |
71cdb697 CH |
2911 | { |
2912 | struct mapped_device *md = bdev->bd_disk->private_data; | |
2913 | const struct pr_ops *ops; | |
71cdb697 | 2914 | fmode_t mode; |
956a4025 | 2915 | int r; |
71cdb697 | 2916 | |
956a4025 | 2917 | r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); |
71cdb697 CH |
2918 | if (r < 0) |
2919 | return r; | |
2920 | ||
2921 | ops = bdev->bd_disk->fops->pr_ops; | |
2922 | if (ops && ops->pr_reserve) | |
2923 | r = ops->pr_reserve(bdev, key, type, flags); | |
2924 | else | |
2925 | r = -EOPNOTSUPP; | |
2926 | ||
956a4025 | 2927 | bdput(bdev); |
71cdb697 CH |
2928 | return r; |
2929 | } | |
2930 | ||
2931 | static int dm_pr_release(struct block_device *bdev, u64 key, enum pr_type type) | |
2932 | { | |
2933 | struct mapped_device *md = bdev->bd_disk->private_data; | |
2934 | const struct pr_ops *ops; | |
71cdb697 | 2935 | fmode_t mode; |
956a4025 | 2936 | int r; |
71cdb697 | 2937 | |
956a4025 | 2938 | r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); |
71cdb697 CH |
2939 | if (r < 0) |
2940 | return r; | |
2941 | ||
2942 | ops = bdev->bd_disk->fops->pr_ops; | |
2943 | if (ops && ops->pr_release) | |
2944 | r = ops->pr_release(bdev, key, type); | |
2945 | else | |
2946 | r = -EOPNOTSUPP; | |
2947 | ||
956a4025 | 2948 | bdput(bdev); |
71cdb697 CH |
2949 | return r; |
2950 | } | |
2951 | ||
2952 | static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key, | |
956a4025 | 2953 | enum pr_type type, bool abort) |
71cdb697 CH |
2954 | { |
2955 | struct mapped_device *md = bdev->bd_disk->private_data; | |
2956 | const struct pr_ops *ops; | |
71cdb697 | 2957 | fmode_t mode; |
956a4025 | 2958 | int r; |
71cdb697 | 2959 | |
956a4025 | 2960 | r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); |
71cdb697 CH |
2961 | if (r < 0) |
2962 | return r; | |
2963 | ||
2964 | ops = bdev->bd_disk->fops->pr_ops; | |
2965 | if (ops && ops->pr_preempt) | |
2966 | r = ops->pr_preempt(bdev, old_key, new_key, type, abort); | |
2967 | else | |
2968 | r = -EOPNOTSUPP; | |
2969 | ||
956a4025 | 2970 | bdput(bdev); |
71cdb697 CH |
2971 | return r; |
2972 | } | |
2973 | ||
2974 | static int dm_pr_clear(struct block_device *bdev, u64 key) | |
2975 | { | |
2976 | struct mapped_device *md = bdev->bd_disk->private_data; | |
2977 | const struct pr_ops *ops; | |
71cdb697 | 2978 | fmode_t mode; |
956a4025 | 2979 | int r; |
71cdb697 | 2980 | |
956a4025 | 2981 | r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); |
71cdb697 CH |
2982 | if (r < 0) |
2983 | return r; | |
2984 | ||
2985 | ops = bdev->bd_disk->fops->pr_ops; | |
2986 | if (ops && ops->pr_clear) | |
2987 | r = ops->pr_clear(bdev, key); | |
2988 | else | |
2989 | r = -EOPNOTSUPP; | |
2990 | ||
956a4025 | 2991 | bdput(bdev); |
71cdb697 CH |
2992 | return r; |
2993 | } | |
2994 | ||
2995 | static const struct pr_ops dm_pr_ops = { | |
2996 | .pr_register = dm_pr_register, | |
2997 | .pr_reserve = dm_pr_reserve, | |
2998 | .pr_release = dm_pr_release, | |
2999 | .pr_preempt = dm_pr_preempt, | |
3000 | .pr_clear = dm_pr_clear, | |
3001 | }; | |
3002 | ||
83d5cde4 | 3003 | static const struct block_device_operations dm_blk_dops = { |
1da177e4 LT |
3004 | .open = dm_blk_open, |
3005 | .release = dm_blk_close, | |
aa129a22 | 3006 | .ioctl = dm_blk_ioctl, |
3ac51e74 | 3007 | .getgeo = dm_blk_getgeo, |
71cdb697 | 3008 | .pr_ops = &dm_pr_ops, |
1da177e4 LT |
3009 | .owner = THIS_MODULE |
3010 | }; | |
3011 | ||
f26c5719 DW |
3012 | static const struct dax_operations dm_dax_ops = { |
3013 | .direct_access = dm_dax_direct_access, | |
7e026c8c | 3014 | .copy_from_iter = dm_dax_copy_from_iter, |
f26c5719 DW |
3015 | }; |
3016 | ||
1da177e4 LT |
3017 | /* |
3018 | * module hooks | |
3019 | */ | |
3020 | module_init(dm_init); | |
3021 | module_exit(dm_exit); | |
3022 | ||
3023 | module_param(major, uint, 0); | |
3024 | MODULE_PARM_DESC(major, "The major number of the device mapper"); | |
f4790826 | 3025 | |
e8603136 MS |
3026 | module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR); |
3027 | MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools"); | |
3028 | ||
115485e8 MS |
3029 | module_param(dm_numa_node, int, S_IRUGO | S_IWUSR); |
3030 | MODULE_PARM_DESC(dm_numa_node, "NUMA node for DM device memory allocations"); | |
3031 | ||
1da177e4 LT |
3032 | MODULE_DESCRIPTION(DM_NAME " driver"); |
3033 | MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>"); | |
3034 | MODULE_LICENSE("GPL"); |