]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (C) 2001 Sistina Software (UK) Limited. | |
d5816876 | 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 | 8 | #include "dm-core.h" |
1da177e4 LT |
9 | |
10 | #include <linux/module.h> | |
11 | #include <linux/vmalloc.h> | |
12 | #include <linux/blkdev.h> | |
13 | #include <linux/namei.h> | |
14 | #include <linux/ctype.h> | |
e7d2860b | 15 | #include <linux/string.h> |
1da177e4 LT |
16 | #include <linux/slab.h> |
17 | #include <linux/interrupt.h> | |
48c9c27b | 18 | #include <linux/mutex.h> |
d5816876 | 19 | #include <linux/delay.h> |
60063497 | 20 | #include <linux/atomic.h> |
bfebd1cd | 21 | #include <linux/blk-mq.h> |
644bda6f | 22 | #include <linux/mount.h> |
1da177e4 | 23 | |
72d94861 AK |
24 | #define DM_MSG_PREFIX "table" |
25 | ||
1da177e4 LT |
26 | #define MAX_DEPTH 16 |
27 | #define NODE_SIZE L1_CACHE_BYTES | |
28 | #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t)) | |
29 | #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1) | |
30 | ||
31 | struct dm_table { | |
1134e5ae | 32 | struct mapped_device *md; |
e6ee8c0b | 33 | unsigned type; |
1da177e4 LT |
34 | |
35 | /* btree table */ | |
36 | unsigned int depth; | |
37 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | |
38 | sector_t *index[MAX_DEPTH]; | |
39 | ||
40 | unsigned int num_targets; | |
41 | unsigned int num_allocated; | |
42 | sector_t *highs; | |
43 | struct dm_target *targets; | |
44 | ||
36a0456f | 45 | struct target_type *immutable_target_type; |
e83068a5 MS |
46 | |
47 | bool integrity_supported:1; | |
48 | bool singleton:1; | |
49 | bool all_blk_mq:1; | |
5ae89a87 | 50 | |
1da177e4 LT |
51 | /* |
52 | * Indicates the rw permissions for the new logical | |
53 | * device. This should be a combination of FMODE_READ | |
54 | * and FMODE_WRITE. | |
55 | */ | |
aeb5d727 | 56 | fmode_t mode; |
1da177e4 LT |
57 | |
58 | /* a list of devices used by this table */ | |
59 | struct list_head devices; | |
60 | ||
1da177e4 LT |
61 | /* events get handed up using this callback */ |
62 | void (*event_fn)(void *); | |
63 | void *event_context; | |
e6ee8c0b KU |
64 | |
65 | struct dm_md_mempools *mempools; | |
9d357b07 N |
66 | |
67 | struct list_head target_callbacks; | |
1da177e4 LT |
68 | }; |
69 | ||
70 | /* | |
71 | * Similar to ceiling(log_size(n)) | |
72 | */ | |
73 | static unsigned int int_log(unsigned int n, unsigned int base) | |
74 | { | |
75 | int result = 0; | |
76 | ||
77 | while (n > 1) { | |
78 | n = dm_div_up(n, base); | |
79 | result++; | |
80 | } | |
81 | ||
82 | return result; | |
83 | } | |
84 | ||
1da177e4 LT |
85 | /* |
86 | * Calculate the index of the child node of the n'th node k'th key. | |
87 | */ | |
88 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
89 | { | |
90 | return (n * CHILDREN_PER_NODE) + k; | |
91 | } | |
92 | ||
93 | /* | |
94 | * Return the n'th node of level l from table t. | |
95 | */ | |
96 | static inline sector_t *get_node(struct dm_table *t, | |
97 | unsigned int l, unsigned int n) | |
98 | { | |
99 | return t->index[l] + (n * KEYS_PER_NODE); | |
100 | } | |
101 | ||
102 | /* | |
103 | * Return the highest key that you could lookup from the n'th | |
104 | * node on level l of the btree. | |
105 | */ | |
106 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
107 | { | |
108 | for (; l < t->depth - 1; l++) | |
109 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
110 | ||
111 | if (n >= t->counts[l]) | |
112 | return (sector_t) - 1; | |
113 | ||
114 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
115 | } | |
116 | ||
117 | /* | |
118 | * Fills in a level of the btree based on the highs of the level | |
119 | * below it. | |
120 | */ | |
121 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
122 | { | |
123 | unsigned int n, k; | |
124 | sector_t *node; | |
125 | ||
126 | for (n = 0U; n < t->counts[l]; n++) { | |
127 | node = get_node(t, l, n); | |
128 | ||
129 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
130 | node[k] = high(t, l + 1, get_child(n, k)); | |
131 | } | |
132 | ||
133 | return 0; | |
134 | } | |
135 | ||
136 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
137 | { | |
138 | unsigned long size; | |
139 | void *addr; | |
140 | ||
141 | /* | |
142 | * Check that we're not going to overflow. | |
143 | */ | |
144 | if (nmemb > (ULONG_MAX / elem_size)) | |
145 | return NULL; | |
146 | ||
147 | size = nmemb * elem_size; | |
e29e65aa | 148 | addr = vzalloc(size); |
1da177e4 LT |
149 | |
150 | return addr; | |
151 | } | |
08649012 | 152 | EXPORT_SYMBOL(dm_vcalloc); |
1da177e4 LT |
153 | |
154 | /* | |
155 | * highs, and targets are managed as dynamic arrays during a | |
156 | * table load. | |
157 | */ | |
158 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
159 | { | |
160 | sector_t *n_highs; | |
161 | struct dm_target *n_targets; | |
1da177e4 LT |
162 | |
163 | /* | |
164 | * Allocate both the target array and offset array at once. | |
512875bd JN |
165 | * Append an empty entry to catch sectors beyond the end of |
166 | * the device. | |
1da177e4 | 167 | */ |
512875bd | 168 | n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) + |
1da177e4 LT |
169 | sizeof(sector_t)); |
170 | if (!n_highs) | |
171 | return -ENOMEM; | |
172 | ||
173 | n_targets = (struct dm_target *) (n_highs + num); | |
174 | ||
57a2f238 | 175 | memset(n_highs, -1, sizeof(*n_highs) * num); |
1da177e4 LT |
176 | vfree(t->highs); |
177 | ||
178 | t->num_allocated = num; | |
179 | t->highs = n_highs; | |
180 | t->targets = n_targets; | |
181 | ||
182 | return 0; | |
183 | } | |
184 | ||
aeb5d727 | 185 | int dm_table_create(struct dm_table **result, fmode_t mode, |
1134e5ae | 186 | unsigned num_targets, struct mapped_device *md) |
1da177e4 | 187 | { |
094262db | 188 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
189 | |
190 | if (!t) | |
191 | return -ENOMEM; | |
192 | ||
1da177e4 | 193 | INIT_LIST_HEAD(&t->devices); |
9d357b07 | 194 | INIT_LIST_HEAD(&t->target_callbacks); |
1da177e4 LT |
195 | |
196 | if (!num_targets) | |
197 | num_targets = KEYS_PER_NODE; | |
198 | ||
199 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
200 | ||
5b2d0657 MP |
201 | if (!num_targets) { |
202 | kfree(t); | |
203 | return -ENOMEM; | |
204 | } | |
205 | ||
1da177e4 LT |
206 | if (alloc_targets(t, num_targets)) { |
207 | kfree(t); | |
1da177e4 LT |
208 | return -ENOMEM; |
209 | } | |
210 | ||
e83068a5 | 211 | t->type = DM_TYPE_NONE; |
1da177e4 | 212 | t->mode = mode; |
1134e5ae | 213 | t->md = md; |
1da177e4 LT |
214 | *result = t; |
215 | return 0; | |
216 | } | |
217 | ||
86f1152b | 218 | static void free_devices(struct list_head *devices, struct mapped_device *md) |
1da177e4 LT |
219 | { |
220 | struct list_head *tmp, *next; | |
221 | ||
afb24528 | 222 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
223 | struct dm_dev_internal *dd = |
224 | list_entry(tmp, struct dm_dev_internal, list); | |
86f1152b BM |
225 | DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s", |
226 | dm_device_name(md), dd->dm_dev->name); | |
227 | dm_put_table_device(md, dd->dm_dev); | |
1da177e4 LT |
228 | kfree(dd); |
229 | } | |
230 | } | |
231 | ||
d5816876 | 232 | void dm_table_destroy(struct dm_table *t) |
1da177e4 LT |
233 | { |
234 | unsigned int i; | |
235 | ||
a7940155 AK |
236 | if (!t) |
237 | return; | |
238 | ||
26803b9f | 239 | /* free the indexes */ |
1da177e4 LT |
240 | if (t->depth >= 2) |
241 | vfree(t->index[t->depth - 2]); | |
242 | ||
243 | /* free the targets */ | |
244 | for (i = 0; i < t->num_targets; i++) { | |
245 | struct dm_target *tgt = t->targets + i; | |
246 | ||
247 | if (tgt->type->dtr) | |
248 | tgt->type->dtr(tgt); | |
249 | ||
250 | dm_put_target_type(tgt->type); | |
251 | } | |
252 | ||
253 | vfree(t->highs); | |
254 | ||
255 | /* free the device list */ | |
86f1152b | 256 | free_devices(&t->devices, t->md); |
1da177e4 | 257 | |
e6ee8c0b KU |
258 | dm_free_md_mempools(t->mempools); |
259 | ||
1da177e4 LT |
260 | kfree(t); |
261 | } | |
262 | ||
1da177e4 LT |
263 | /* |
264 | * See if we've already got a device in the list. | |
265 | */ | |
82b1519b | 266 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 267 | { |
82b1519b | 268 | struct dm_dev_internal *dd; |
1da177e4 LT |
269 | |
270 | list_for_each_entry (dd, l, list) | |
86f1152b | 271 | if (dd->dm_dev->bdev->bd_dev == dev) |
1da177e4 LT |
272 | return dd; |
273 | ||
274 | return NULL; | |
275 | } | |
276 | ||
1da177e4 | 277 | /* |
f6a1ed10 | 278 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 279 | */ |
f6a1ed10 MP |
280 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
281 | sector_t start, sector_t len, void *data) | |
1da177e4 | 282 | { |
f4808ca9 | 283 | struct request_queue *q; |
754c5fc7 MS |
284 | struct queue_limits *limits = data; |
285 | struct block_device *bdev = dev->bdev; | |
286 | sector_t dev_size = | |
287 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 288 | unsigned short logical_block_size_sectors = |
754c5fc7 | 289 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 290 | char b[BDEVNAME_SIZE]; |
2cd54d9b | 291 | |
f4808ca9 MB |
292 | /* |
293 | * Some devices exist without request functions, | |
294 | * such as loop devices not yet bound to backing files. | |
295 | * Forbid the use of such devices. | |
296 | */ | |
297 | q = bdev_get_queue(bdev); | |
298 | if (!q || !q->make_request_fn) { | |
299 | DMWARN("%s: %s is not yet initialised: " | |
300 | "start=%llu, len=%llu, dev_size=%llu", | |
301 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
302 | (unsigned long long)start, | |
303 | (unsigned long long)len, | |
304 | (unsigned long long)dev_size); | |
305 | return 1; | |
306 | } | |
307 | ||
2cd54d9b | 308 | if (!dev_size) |
f6a1ed10 | 309 | return 0; |
2cd54d9b | 310 | |
5dea271b | 311 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
312 | DMWARN("%s: %s too small for target: " |
313 | "start=%llu, len=%llu, dev_size=%llu", | |
314 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
315 | (unsigned long long)start, | |
316 | (unsigned long long)len, | |
317 | (unsigned long long)dev_size); | |
f6a1ed10 | 318 | return 1; |
02acc3a4 MS |
319 | } |
320 | ||
321 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 322 | return 0; |
02acc3a4 MS |
323 | |
324 | if (start & (logical_block_size_sectors - 1)) { | |
325 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 326 | "logical block size %u of %s", |
02acc3a4 MS |
327 | dm_device_name(ti->table->md), |
328 | (unsigned long long)start, | |
754c5fc7 | 329 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 330 | return 1; |
02acc3a4 MS |
331 | } |
332 | ||
5dea271b | 333 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 334 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 335 | "logical block size %u of %s", |
02acc3a4 | 336 | dm_device_name(ti->table->md), |
5dea271b | 337 | (unsigned long long)len, |
754c5fc7 | 338 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 339 | return 1; |
02acc3a4 MS |
340 | } |
341 | ||
f6a1ed10 | 342 | return 0; |
1da177e4 LT |
343 | } |
344 | ||
345 | /* | |
570b9d96 | 346 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 347 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
348 | * device and not to touch the existing bdev field in case |
349 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 350 | */ |
aeb5d727 | 351 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 352 | struct mapped_device *md) |
1da177e4 LT |
353 | { |
354 | int r; | |
86f1152b | 355 | struct dm_dev *old_dev, *new_dev; |
1da177e4 | 356 | |
86f1152b | 357 | old_dev = dd->dm_dev; |
570b9d96 | 358 | |
86f1152b BM |
359 | r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, |
360 | dd->dm_dev->mode | new_mode, &new_dev); | |
570b9d96 AK |
361 | if (r) |
362 | return r; | |
1da177e4 | 363 | |
86f1152b BM |
364 | dd->dm_dev = new_dev; |
365 | dm_put_table_device(md, old_dev); | |
1da177e4 | 366 | |
570b9d96 | 367 | return 0; |
1da177e4 LT |
368 | } |
369 | ||
4df2bf46 D |
370 | /* |
371 | * Convert the path to a device | |
372 | */ | |
373 | dev_t dm_get_dev_t(const char *path) | |
374 | { | |
375 | dev_t uninitialized_var(dev); | |
376 | struct block_device *bdev; | |
377 | ||
378 | bdev = lookup_bdev(path); | |
379 | if (IS_ERR(bdev)) | |
380 | dev = name_to_dev_t(path); | |
381 | else { | |
382 | dev = bdev->bd_dev; | |
383 | bdput(bdev); | |
384 | } | |
385 | ||
386 | return dev; | |
387 | } | |
388 | EXPORT_SYMBOL_GPL(dm_get_dev_t); | |
389 | ||
1da177e4 LT |
390 | /* |
391 | * Add a device to the list, or just increment the usage count if | |
392 | * it's already present. | |
393 | */ | |
08649012 MS |
394 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
395 | struct dm_dev **result) | |
1da177e4 LT |
396 | { |
397 | int r; | |
4df2bf46 | 398 | dev_t dev; |
82b1519b | 399 | struct dm_dev_internal *dd; |
08649012 | 400 | struct dm_table *t = ti->table; |
1da177e4 | 401 | |
547bc926 | 402 | BUG_ON(!t); |
1da177e4 | 403 | |
4df2bf46 D |
404 | dev = dm_get_dev_t(path); |
405 | if (!dev) | |
406 | return -ENODEV; | |
1da177e4 LT |
407 | |
408 | dd = find_device(&t->devices, dev); | |
409 | if (!dd) { | |
410 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
411 | if (!dd) | |
412 | return -ENOMEM; | |
413 | ||
86f1152b | 414 | if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { |
1da177e4 LT |
415 | kfree(dd); |
416 | return r; | |
417 | } | |
418 | ||
1da177e4 LT |
419 | atomic_set(&dd->count, 0); |
420 | list_add(&dd->list, &t->devices); | |
421 | ||
86f1152b | 422 | } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { |
f165921d | 423 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
424 | if (r) |
425 | return r; | |
426 | } | |
427 | atomic_inc(&dd->count); | |
428 | ||
86f1152b | 429 | *result = dd->dm_dev; |
1da177e4 LT |
430 | return 0; |
431 | } | |
08649012 | 432 | EXPORT_SYMBOL(dm_get_device); |
1da177e4 | 433 | |
11f0431b MS |
434 | static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
435 | sector_t start, sector_t len, void *data) | |
1da177e4 | 436 | { |
754c5fc7 MS |
437 | struct queue_limits *limits = data; |
438 | struct block_device *bdev = dev->bdev; | |
165125e1 | 439 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
440 | char b[BDEVNAME_SIZE]; |
441 | ||
442 | if (unlikely(!q)) { | |
443 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
444 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 445 | return 0; |
0c2322e4 | 446 | } |
3cb40214 | 447 | |
b27d7f16 MP |
448 | if (bdev_stack_limits(limits, bdev, start) < 0) |
449 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
450 | "physical_block_size=%u, logical_block_size=%u, " |
451 | "alignment_offset=%u, start=%llu", | |
452 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
453 | q->limits.physical_block_size, | |
454 | q->limits.logical_block_size, | |
455 | q->limits.alignment_offset, | |
b27d7f16 | 456 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 457 | |
754c5fc7 | 458 | return 0; |
3cb40214 | 459 | } |
969429b5 | 460 | |
1da177e4 | 461 | /* |
08649012 | 462 | * Decrement a device's use count and remove it if necessary. |
1da177e4 | 463 | */ |
82b1519b | 464 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 465 | { |
86f1152b BM |
466 | int found = 0; |
467 | struct list_head *devices = &ti->table->devices; | |
468 | struct dm_dev_internal *dd; | |
82b1519b | 469 | |
86f1152b BM |
470 | list_for_each_entry(dd, devices, list) { |
471 | if (dd->dm_dev == d) { | |
472 | found = 1; | |
473 | break; | |
474 | } | |
475 | } | |
476 | if (!found) { | |
477 | DMWARN("%s: device %s not in table devices list", | |
478 | dm_device_name(ti->table->md), d->name); | |
479 | return; | |
480 | } | |
1da177e4 | 481 | if (atomic_dec_and_test(&dd->count)) { |
86f1152b | 482 | dm_put_table_device(ti->table->md, d); |
1da177e4 LT |
483 | list_del(&dd->list); |
484 | kfree(dd); | |
485 | } | |
486 | } | |
08649012 | 487 | EXPORT_SYMBOL(dm_put_device); |
1da177e4 LT |
488 | |
489 | /* | |
490 | * Checks to see if the target joins onto the end of the table. | |
491 | */ | |
492 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
493 | { | |
494 | struct dm_target *prev; | |
495 | ||
496 | if (!table->num_targets) | |
497 | return !ti->begin; | |
498 | ||
499 | prev = &table->targets[table->num_targets - 1]; | |
500 | return (ti->begin == (prev->begin + prev->len)); | |
501 | } | |
502 | ||
503 | /* | |
504 | * Used to dynamically allocate the arg array. | |
f36afb39 MP |
505 | * |
506 | * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must | |
507 | * process messages even if some device is suspended. These messages have a | |
508 | * small fixed number of arguments. | |
509 | * | |
510 | * On the other hand, dm-switch needs to process bulk data using messages and | |
511 | * excessive use of GFP_NOIO could cause trouble. | |
1da177e4 LT |
512 | */ |
513 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
514 | { | |
515 | char **argv; | |
516 | unsigned new_size; | |
f36afb39 | 517 | gfp_t gfp; |
1da177e4 | 518 | |
f36afb39 MP |
519 | if (*array_size) { |
520 | new_size = *array_size * 2; | |
521 | gfp = GFP_KERNEL; | |
522 | } else { | |
523 | new_size = 8; | |
524 | gfp = GFP_NOIO; | |
525 | } | |
526 | argv = kmalloc(new_size * sizeof(*argv), gfp); | |
1da177e4 LT |
527 | if (argv) { |
528 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
529 | *array_size = new_size; | |
530 | } | |
531 | ||
532 | kfree(old_argv); | |
533 | return argv; | |
534 | } | |
535 | ||
536 | /* | |
537 | * Destructively splits up the argument list to pass to ctr. | |
538 | */ | |
539 | int dm_split_args(int *argc, char ***argvp, char *input) | |
540 | { | |
541 | char *start, *end = input, *out, **argv = NULL; | |
542 | unsigned array_size = 0; | |
543 | ||
544 | *argc = 0; | |
814d6862 DT |
545 | |
546 | if (!input) { | |
547 | *argvp = NULL; | |
548 | return 0; | |
549 | } | |
550 | ||
1da177e4 LT |
551 | argv = realloc_argv(&array_size, argv); |
552 | if (!argv) | |
553 | return -ENOMEM; | |
554 | ||
555 | while (1) { | |
1da177e4 | 556 | /* Skip whitespace */ |
e7d2860b | 557 | start = skip_spaces(end); |
1da177e4 LT |
558 | |
559 | if (!*start) | |
560 | break; /* success, we hit the end */ | |
561 | ||
562 | /* 'out' is used to remove any back-quotes */ | |
563 | end = out = start; | |
564 | while (*end) { | |
565 | /* Everything apart from '\0' can be quoted */ | |
566 | if (*end == '\\' && *(end + 1)) { | |
567 | *out++ = *(end + 1); | |
568 | end += 2; | |
569 | continue; | |
570 | } | |
571 | ||
572 | if (isspace(*end)) | |
573 | break; /* end of token */ | |
574 | ||
575 | *out++ = *end++; | |
576 | } | |
577 | ||
578 | /* have we already filled the array ? */ | |
579 | if ((*argc + 1) > array_size) { | |
580 | argv = realloc_argv(&array_size, argv); | |
581 | if (!argv) | |
582 | return -ENOMEM; | |
583 | } | |
584 | ||
585 | /* we know this is whitespace */ | |
586 | if (*end) | |
587 | end++; | |
588 | ||
589 | /* terminate the string and put it in the array */ | |
590 | *out = '\0'; | |
591 | argv[*argc] = start; | |
592 | (*argc)++; | |
593 | } | |
594 | ||
595 | *argvp = argv; | |
596 | return 0; | |
597 | } | |
598 | ||
be6d4305 MS |
599 | /* |
600 | * Impose necessary and sufficient conditions on a devices's table such | |
601 | * that any incoming bio which respects its logical_block_size can be | |
602 | * processed successfully. If it falls across the boundary between | |
603 | * two or more targets, the size of each piece it gets split into must | |
604 | * be compatible with the logical_block_size of the target processing it. | |
605 | */ | |
754c5fc7 MS |
606 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
607 | struct queue_limits *limits) | |
be6d4305 MS |
608 | { |
609 | /* | |
610 | * This function uses arithmetic modulo the logical_block_size | |
611 | * (in units of 512-byte sectors). | |
612 | */ | |
613 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 614 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
615 | |
616 | /* | |
617 | * Offset of the start of the next table entry, mod logical_block_size. | |
618 | */ | |
619 | unsigned short next_target_start = 0; | |
620 | ||
621 | /* | |
622 | * Given an aligned bio that extends beyond the end of a | |
623 | * target, how many sectors must the next target handle? | |
624 | */ | |
625 | unsigned short remaining = 0; | |
626 | ||
627 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 628 | struct queue_limits ti_limits; |
be6d4305 MS |
629 | unsigned i = 0; |
630 | ||
631 | /* | |
632 | * Check each entry in the table in turn. | |
633 | */ | |
634 | while (i < dm_table_get_num_targets(table)) { | |
635 | ti = dm_table_get_target(table, i++); | |
636 | ||
b1bd055d | 637 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
638 | |
639 | /* combine all target devices' limits */ | |
640 | if (ti->type->iterate_devices) | |
641 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
642 | &ti_limits); | |
643 | ||
be6d4305 MS |
644 | /* |
645 | * If the remaining sectors fall entirely within this | |
646 | * table entry are they compatible with its logical_block_size? | |
647 | */ | |
648 | if (remaining < ti->len && | |
754c5fc7 | 649 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
650 | SECTOR_SHIFT) - 1)) |
651 | break; /* Error */ | |
652 | ||
653 | next_target_start = | |
654 | (unsigned short) ((next_target_start + ti->len) & | |
655 | (device_logical_block_size_sects - 1)); | |
656 | remaining = next_target_start ? | |
657 | device_logical_block_size_sects - next_target_start : 0; | |
658 | } | |
659 | ||
660 | if (remaining) { | |
661 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 662 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
663 | dm_device_name(table->md), i, |
664 | (unsigned long long) ti->begin, | |
665 | (unsigned long long) ti->len, | |
754c5fc7 | 666 | limits->logical_block_size); |
be6d4305 MS |
667 | return -EINVAL; |
668 | } | |
669 | ||
670 | return 0; | |
671 | } | |
672 | ||
1da177e4 LT |
673 | int dm_table_add_target(struct dm_table *t, const char *type, |
674 | sector_t start, sector_t len, char *params) | |
675 | { | |
676 | int r = -EINVAL, argc; | |
677 | char **argv; | |
678 | struct dm_target *tgt; | |
679 | ||
3791e2fc AK |
680 | if (t->singleton) { |
681 | DMERR("%s: target type %s must appear alone in table", | |
682 | dm_device_name(t->md), t->targets->type->name); | |
683 | return -EINVAL; | |
684 | } | |
685 | ||
57a2f238 | 686 | BUG_ON(t->num_targets >= t->num_allocated); |
1da177e4 LT |
687 | |
688 | tgt = t->targets + t->num_targets; | |
689 | memset(tgt, 0, sizeof(*tgt)); | |
690 | ||
691 | if (!len) { | |
72d94861 | 692 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
693 | return -EINVAL; |
694 | } | |
695 | ||
696 | tgt->type = dm_get_target_type(type); | |
697 | if (!tgt->type) { | |
dafa724b | 698 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), type); |
1da177e4 LT |
699 | return -EINVAL; |
700 | } | |
701 | ||
3791e2fc AK |
702 | if (dm_target_needs_singleton(tgt->type)) { |
703 | if (t->num_targets) { | |
dafa724b | 704 | tgt->error = "singleton target type must appear alone in table"; |
705 | goto bad; | |
3791e2fc | 706 | } |
e83068a5 | 707 | t->singleton = true; |
3791e2fc AK |
708 | } |
709 | ||
cc6cbe14 | 710 | if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) { |
dafa724b | 711 | tgt->error = "target type may not be included in a read-only table"; |
712 | goto bad; | |
cc6cbe14 AK |
713 | } |
714 | ||
36a0456f AK |
715 | if (t->immutable_target_type) { |
716 | if (t->immutable_target_type != tgt->type) { | |
dafa724b | 717 | tgt->error = "immutable target type cannot be mixed with other target types"; |
718 | goto bad; | |
36a0456f AK |
719 | } |
720 | } else if (dm_target_is_immutable(tgt->type)) { | |
721 | if (t->num_targets) { | |
dafa724b | 722 | tgt->error = "immutable target type cannot be mixed with other target types"; |
723 | goto bad; | |
36a0456f AK |
724 | } |
725 | t->immutable_target_type = tgt->type; | |
726 | } | |
727 | ||
1da177e4 LT |
728 | tgt->table = t; |
729 | tgt->begin = start; | |
730 | tgt->len = len; | |
731 | tgt->error = "Unknown error"; | |
732 | ||
733 | /* | |
734 | * Does this target adjoin the previous one ? | |
735 | */ | |
736 | if (!adjoin(t, tgt)) { | |
737 | tgt->error = "Gap in table"; | |
1da177e4 LT |
738 | goto bad; |
739 | } | |
740 | ||
741 | r = dm_split_args(&argc, &argv, params); | |
742 | if (r) { | |
743 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
744 | goto bad; | |
745 | } | |
746 | ||
747 | r = tgt->type->ctr(tgt, argc, argv); | |
748 | kfree(argv); | |
749 | if (r) | |
750 | goto bad; | |
751 | ||
752 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
753 | ||
55a62eef AK |
754 | if (!tgt->num_discard_bios && tgt->discards_supported) |
755 | DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.", | |
936688d7 | 756 | dm_device_name(t->md), type); |
5ae89a87 | 757 | |
1da177e4 LT |
758 | return 0; |
759 | ||
760 | bad: | |
72d94861 | 761 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
762 | dm_put_target_type(tgt->type); |
763 | return r; | |
764 | } | |
765 | ||
498f0103 MS |
766 | /* |
767 | * Target argument parsing helpers. | |
768 | */ | |
769 | static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
770 | unsigned *value, char **error, unsigned grouped) | |
771 | { | |
772 | const char *arg_str = dm_shift_arg(arg_set); | |
31998ef1 | 773 | char dummy; |
498f0103 MS |
774 | |
775 | if (!arg_str || | |
31998ef1 | 776 | (sscanf(arg_str, "%u%c", value, &dummy) != 1) || |
498f0103 MS |
777 | (*value < arg->min) || |
778 | (*value > arg->max) || | |
779 | (grouped && arg_set->argc < *value)) { | |
780 | *error = arg->error; | |
781 | return -EINVAL; | |
782 | } | |
783 | ||
784 | return 0; | |
785 | } | |
786 | ||
787 | int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
788 | unsigned *value, char **error) | |
789 | { | |
790 | return validate_next_arg(arg, arg_set, value, error, 0); | |
791 | } | |
792 | EXPORT_SYMBOL(dm_read_arg); | |
793 | ||
794 | int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
795 | unsigned *value, char **error) | |
796 | { | |
797 | return validate_next_arg(arg, arg_set, value, error, 1); | |
798 | } | |
799 | EXPORT_SYMBOL(dm_read_arg_group); | |
800 | ||
801 | const char *dm_shift_arg(struct dm_arg_set *as) | |
802 | { | |
803 | char *r; | |
804 | ||
805 | if (as->argc) { | |
806 | as->argc--; | |
807 | r = *as->argv; | |
808 | as->argv++; | |
809 | return r; | |
810 | } | |
811 | ||
812 | return NULL; | |
813 | } | |
814 | EXPORT_SYMBOL(dm_shift_arg); | |
815 | ||
816 | void dm_consume_args(struct dm_arg_set *as, unsigned num_args) | |
817 | { | |
818 | BUG_ON(as->argc < num_args); | |
819 | as->argc -= num_args; | |
820 | as->argv += num_args; | |
821 | } | |
822 | EXPORT_SYMBOL(dm_consume_args); | |
823 | ||
545ed20e TK |
824 | static bool __table_type_bio_based(unsigned table_type) |
825 | { | |
826 | return (table_type == DM_TYPE_BIO_BASED || | |
827 | table_type == DM_TYPE_DAX_BIO_BASED); | |
828 | } | |
829 | ||
15b94a69 JN |
830 | static bool __table_type_request_based(unsigned table_type) |
831 | { | |
832 | return (table_type == DM_TYPE_REQUEST_BASED || | |
833 | table_type == DM_TYPE_MQ_REQUEST_BASED); | |
834 | } | |
835 | ||
e83068a5 MS |
836 | void dm_table_set_type(struct dm_table *t, unsigned type) |
837 | { | |
838 | t->type = type; | |
839 | } | |
840 | EXPORT_SYMBOL_GPL(dm_table_set_type); | |
841 | ||
545ed20e TK |
842 | static int device_supports_dax(struct dm_target *ti, struct dm_dev *dev, |
843 | sector_t start, sector_t len, void *data) | |
844 | { | |
845 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
846 | ||
847 | return q && blk_queue_dax(q); | |
848 | } | |
849 | ||
850 | static bool dm_table_supports_dax(struct dm_table *t) | |
851 | { | |
852 | struct dm_target *ti; | |
853 | unsigned i = 0; | |
854 | ||
855 | /* Ensure that all targets support DAX. */ | |
856 | while (i < dm_table_get_num_targets(t)) { | |
857 | ti = dm_table_get_target(t, i++); | |
858 | ||
859 | if (!ti->type->direct_access) | |
860 | return false; | |
861 | ||
862 | if (!ti->type->iterate_devices || | |
863 | !ti->type->iterate_devices(ti, device_supports_dax, NULL)) | |
864 | return false; | |
865 | } | |
866 | ||
867 | return true; | |
868 | } | |
869 | ||
e83068a5 | 870 | static int dm_table_determine_type(struct dm_table *t) |
e6ee8c0b KU |
871 | { |
872 | unsigned i; | |
169e2cc2 | 873 | unsigned bio_based = 0, request_based = 0, hybrid = 0; |
5b8c01f7 | 874 | unsigned sq_count = 0, mq_count = 0; |
e6ee8c0b KU |
875 | struct dm_target *tgt; |
876 | struct dm_dev_internal *dd; | |
e83068a5 | 877 | struct list_head *devices = dm_table_get_devices(t); |
65803c20 | 878 | unsigned live_md_type = dm_get_md_type(t->md); |
e6ee8c0b | 879 | |
e83068a5 MS |
880 | if (t->type != DM_TYPE_NONE) { |
881 | /* target already set the table's type */ | |
882 | if (t->type == DM_TYPE_BIO_BASED) | |
883 | return 0; | |
545ed20e | 884 | BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED); |
e83068a5 MS |
885 | goto verify_rq_based; |
886 | } | |
887 | ||
e6ee8c0b KU |
888 | for (i = 0; i < t->num_targets; i++) { |
889 | tgt = t->targets + i; | |
169e2cc2 MS |
890 | if (dm_target_hybrid(tgt)) |
891 | hybrid = 1; | |
892 | else if (dm_target_request_based(tgt)) | |
e6ee8c0b KU |
893 | request_based = 1; |
894 | else | |
895 | bio_based = 1; | |
896 | ||
897 | if (bio_based && request_based) { | |
898 | DMWARN("Inconsistent table: different target types" | |
899 | " can't be mixed up"); | |
900 | return -EINVAL; | |
901 | } | |
902 | } | |
903 | ||
169e2cc2 MS |
904 | if (hybrid && !bio_based && !request_based) { |
905 | /* | |
906 | * The targets can work either way. | |
907 | * Determine the type from the live device. | |
908 | * Default to bio-based if device is new. | |
909 | */ | |
15b94a69 | 910 | if (__table_type_request_based(live_md_type)) |
169e2cc2 MS |
911 | request_based = 1; |
912 | else | |
913 | bio_based = 1; | |
914 | } | |
915 | ||
e6ee8c0b KU |
916 | if (bio_based) { |
917 | /* We must use this table as bio-based */ | |
918 | t->type = DM_TYPE_BIO_BASED; | |
f8df1fdf MS |
919 | if (dm_table_supports_dax(t) || |
920 | (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) | |
545ed20e | 921 | t->type = DM_TYPE_DAX_BIO_BASED; |
e6ee8c0b KU |
922 | return 0; |
923 | } | |
924 | ||
925 | BUG_ON(!request_based); /* No targets in this table */ | |
926 | ||
e83068a5 MS |
927 | /* |
928 | * The only way to establish DM_TYPE_MQ_REQUEST_BASED is by | |
929 | * having a compatible target use dm_table_set_type. | |
930 | */ | |
931 | t->type = DM_TYPE_REQUEST_BASED; | |
932 | ||
933 | verify_rq_based: | |
65803c20 MS |
934 | /* |
935 | * Request-based dm supports only tables that have a single target now. | |
936 | * To support multiple targets, request splitting support is needed, | |
937 | * and that needs lots of changes in the block-layer. | |
938 | * (e.g. request completion process for partial completion.) | |
939 | */ | |
940 | if (t->num_targets > 1) { | |
941 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
942 | return -EINVAL; | |
943 | } | |
944 | ||
6936c12c MS |
945 | if (list_empty(devices)) { |
946 | int srcu_idx; | |
947 | struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx); | |
948 | ||
949 | /* inherit live table's type and all_blk_mq */ | |
950 | if (live_table) { | |
951 | t->type = live_table->type; | |
952 | t->all_blk_mq = live_table->all_blk_mq; | |
953 | } | |
954 | dm_put_live_table(t->md, srcu_idx); | |
955 | return 0; | |
956 | } | |
957 | ||
e6ee8c0b | 958 | /* Non-request-stackable devices can't be used for request-based dm */ |
e6ee8c0b | 959 | list_for_each_entry(dd, devices, list) { |
e5863d9a MS |
960 | struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev); |
961 | ||
962 | if (!blk_queue_stackable(q)) { | |
963 | DMERR("table load rejected: including" | |
964 | " non-request-stackable devices"); | |
e6ee8c0b KU |
965 | return -EINVAL; |
966 | } | |
e5863d9a MS |
967 | |
968 | if (q->mq_ops) | |
5b8c01f7 BVA |
969 | mq_count++; |
970 | else | |
971 | sq_count++; | |
e5863d9a | 972 | } |
5b8c01f7 BVA |
973 | if (sq_count && mq_count) { |
974 | DMERR("table load rejected: not all devices are blk-mq request-stackable"); | |
975 | return -EINVAL; | |
e83068a5 | 976 | } |
5b8c01f7 | 977 | t->all_blk_mq = mq_count > 0; |
e6ee8c0b | 978 | |
301fc3f5 BVA |
979 | if (t->type == DM_TYPE_MQ_REQUEST_BASED && !t->all_blk_mq) { |
980 | DMERR("table load rejected: all devices are not blk-mq request-stackable"); | |
981 | return -EINVAL; | |
982 | } | |
983 | ||
e6ee8c0b KU |
984 | return 0; |
985 | } | |
986 | ||
987 | unsigned dm_table_get_type(struct dm_table *t) | |
988 | { | |
989 | return t->type; | |
990 | } | |
991 | ||
36a0456f AK |
992 | struct target_type *dm_table_get_immutable_target_type(struct dm_table *t) |
993 | { | |
994 | return t->immutable_target_type; | |
995 | } | |
996 | ||
16f12266 MS |
997 | struct dm_target *dm_table_get_immutable_target(struct dm_table *t) |
998 | { | |
999 | /* Immutable target is implicitly a singleton */ | |
1000 | if (t->num_targets > 1 || | |
1001 | !dm_target_is_immutable(t->targets[0].type)) | |
1002 | return NULL; | |
1003 | ||
1004 | return t->targets; | |
1005 | } | |
1006 | ||
f083b09b MS |
1007 | struct dm_target *dm_table_get_wildcard_target(struct dm_table *t) |
1008 | { | |
1009 | struct dm_target *uninitialized_var(ti); | |
1010 | unsigned i = 0; | |
1011 | ||
1012 | while (i < dm_table_get_num_targets(t)) { | |
1013 | ti = dm_table_get_target(t, i++); | |
1014 | if (dm_target_is_wildcard(ti->type)) | |
1015 | return ti; | |
1016 | } | |
1017 | ||
1018 | return NULL; | |
1019 | } | |
1020 | ||
545ed20e TK |
1021 | bool dm_table_bio_based(struct dm_table *t) |
1022 | { | |
1023 | return __table_type_bio_based(dm_table_get_type(t)); | |
1024 | } | |
1025 | ||
e6ee8c0b KU |
1026 | bool dm_table_request_based(struct dm_table *t) |
1027 | { | |
15b94a69 | 1028 | return __table_type_request_based(dm_table_get_type(t)); |
e5863d9a MS |
1029 | } |
1030 | ||
e83068a5 | 1031 | bool dm_table_all_blk_mq_devices(struct dm_table *t) |
e5863d9a | 1032 | { |
e83068a5 | 1033 | return t->all_blk_mq; |
e6ee8c0b KU |
1034 | } |
1035 | ||
17e149b8 | 1036 | static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md) |
e6ee8c0b KU |
1037 | { |
1038 | unsigned type = dm_table_get_type(t); | |
30187e1d | 1039 | unsigned per_io_data_size = 0; |
78d8e58a | 1040 | struct dm_target *tgt; |
c0820cf5 | 1041 | unsigned i; |
e6ee8c0b | 1042 | |
78d8e58a | 1043 | if (unlikely(type == DM_TYPE_NONE)) { |
e6ee8c0b KU |
1044 | DMWARN("no table type is set, can't allocate mempools"); |
1045 | return -EINVAL; | |
1046 | } | |
1047 | ||
545ed20e | 1048 | if (__table_type_bio_based(type)) |
78d8e58a MS |
1049 | for (i = 0; i < t->num_targets; i++) { |
1050 | tgt = t->targets + i; | |
30187e1d | 1051 | per_io_data_size = max(per_io_data_size, tgt->per_io_data_size); |
78d8e58a MS |
1052 | } |
1053 | ||
30187e1d | 1054 | t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported, per_io_data_size); |
4e6e36c3 MS |
1055 | if (!t->mempools) |
1056 | return -ENOMEM; | |
e6ee8c0b KU |
1057 | |
1058 | return 0; | |
1059 | } | |
1060 | ||
1061 | void dm_table_free_md_mempools(struct dm_table *t) | |
1062 | { | |
1063 | dm_free_md_mempools(t->mempools); | |
1064 | t->mempools = NULL; | |
1065 | } | |
1066 | ||
1067 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
1068 | { | |
1069 | return t->mempools; | |
1070 | } | |
1071 | ||
1da177e4 LT |
1072 | static int setup_indexes(struct dm_table *t) |
1073 | { | |
1074 | int i; | |
1075 | unsigned int total = 0; | |
1076 | sector_t *indexes; | |
1077 | ||
1078 | /* allocate the space for *all* the indexes */ | |
1079 | for (i = t->depth - 2; i >= 0; i--) { | |
1080 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
1081 | total += t->counts[i]; | |
1082 | } | |
1083 | ||
1084 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
1085 | if (!indexes) | |
1086 | return -ENOMEM; | |
1087 | ||
1088 | /* set up internal nodes, bottom-up */ | |
82d601dc | 1089 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
1090 | t->index[i] = indexes; |
1091 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
1092 | setup_btree_index(i, t); | |
1093 | } | |
1094 | ||
1095 | return 0; | |
1096 | } | |
1097 | ||
1098 | /* | |
1099 | * Builds the btree to index the map. | |
1100 | */ | |
26803b9f | 1101 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
1102 | { |
1103 | int r = 0; | |
1104 | unsigned int leaf_nodes; | |
1105 | ||
1da177e4 LT |
1106 | /* how many indexes will the btree have ? */ |
1107 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
1108 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
1109 | ||
1110 | /* leaf layer has already been set up */ | |
1111 | t->counts[t->depth - 1] = leaf_nodes; | |
1112 | t->index[t->depth - 1] = t->highs; | |
1113 | ||
1114 | if (t->depth >= 2) | |
1115 | r = setup_indexes(t); | |
1116 | ||
1117 | return r; | |
1118 | } | |
1119 | ||
25520d55 MP |
1120 | static bool integrity_profile_exists(struct gendisk *disk) |
1121 | { | |
1122 | return !!blk_get_integrity(disk); | |
1123 | } | |
1124 | ||
a63a5cf8 MS |
1125 | /* |
1126 | * Get a disk whose integrity profile reflects the table's profile. | |
a63a5cf8 MS |
1127 | * Returns NULL if integrity support was inconsistent or unavailable. |
1128 | */ | |
25520d55 | 1129 | static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t) |
a63a5cf8 MS |
1130 | { |
1131 | struct list_head *devices = dm_table_get_devices(t); | |
1132 | struct dm_dev_internal *dd = NULL; | |
1133 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
1134 | ||
1135 | list_for_each_entry(dd, devices, list) { | |
86f1152b | 1136 | template_disk = dd->dm_dev->bdev->bd_disk; |
25520d55 | 1137 | if (!integrity_profile_exists(template_disk)) |
a63a5cf8 | 1138 | goto no_integrity; |
a63a5cf8 MS |
1139 | else if (prev_disk && |
1140 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
1141 | goto no_integrity; | |
1142 | prev_disk = template_disk; | |
1143 | } | |
1144 | ||
1145 | return template_disk; | |
1146 | ||
1147 | no_integrity: | |
1148 | if (prev_disk) | |
1149 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
1150 | dm_device_name(t->md), | |
1151 | prev_disk->disk_name, | |
1152 | template_disk->disk_name); | |
1153 | return NULL; | |
1154 | } | |
1155 | ||
26803b9f | 1156 | /* |
25520d55 MP |
1157 | * Register the mapped device for blk_integrity support if the |
1158 | * underlying devices have an integrity profile. But all devices may | |
1159 | * not have matching profiles (checking all devices isn't reliable | |
a63a5cf8 | 1160 | * during table load because this table may use other DM device(s) which |
25520d55 MP |
1161 | * must be resumed before they will have an initialized integity |
1162 | * profile). Consequently, stacked DM devices force a 2 stage integrity | |
1163 | * profile validation: First pass during table load, final pass during | |
1164 | * resume. | |
26803b9f | 1165 | */ |
25520d55 | 1166 | static int dm_table_register_integrity(struct dm_table *t) |
26803b9f | 1167 | { |
25520d55 | 1168 | struct mapped_device *md = t->md; |
a63a5cf8 | 1169 | struct gendisk *template_disk = NULL; |
26803b9f | 1170 | |
25520d55 | 1171 | template_disk = dm_table_get_integrity_disk(t); |
a63a5cf8 MS |
1172 | if (!template_disk) |
1173 | return 0; | |
26803b9f | 1174 | |
25520d55 | 1175 | if (!integrity_profile_exists(dm_disk(md))) { |
e83068a5 | 1176 | t->integrity_supported = true; |
25520d55 MP |
1177 | /* |
1178 | * Register integrity profile during table load; we can do | |
1179 | * this because the final profile must match during resume. | |
1180 | */ | |
1181 | blk_integrity_register(dm_disk(md), | |
1182 | blk_get_integrity(template_disk)); | |
1183 | return 0; | |
a63a5cf8 MS |
1184 | } |
1185 | ||
1186 | /* | |
25520d55 | 1187 | * If DM device already has an initialized integrity |
a63a5cf8 MS |
1188 | * profile the new profile should not conflict. |
1189 | */ | |
25520d55 | 1190 | if (blk_integrity_compare(dm_disk(md), template_disk) < 0) { |
a63a5cf8 MS |
1191 | DMWARN("%s: conflict with existing integrity profile: " |
1192 | "%s profile mismatch", | |
1193 | dm_device_name(t->md), | |
1194 | template_disk->disk_name); | |
1195 | return 1; | |
1196 | } | |
1197 | ||
25520d55 | 1198 | /* Preserve existing integrity profile */ |
e83068a5 | 1199 | t->integrity_supported = true; |
26803b9f WD |
1200 | return 0; |
1201 | } | |
1202 | ||
1203 | /* | |
1204 | * Prepares the table for use by building the indices, | |
1205 | * setting the type, and allocating mempools. | |
1206 | */ | |
1207 | int dm_table_complete(struct dm_table *t) | |
1208 | { | |
1209 | int r; | |
1210 | ||
e83068a5 | 1211 | r = dm_table_determine_type(t); |
26803b9f | 1212 | if (r) { |
e83068a5 | 1213 | DMERR("unable to determine table type"); |
26803b9f WD |
1214 | return r; |
1215 | } | |
1216 | ||
1217 | r = dm_table_build_index(t); | |
1218 | if (r) { | |
1219 | DMERR("unable to build btrees"); | |
1220 | return r; | |
1221 | } | |
1222 | ||
25520d55 | 1223 | r = dm_table_register_integrity(t); |
26803b9f WD |
1224 | if (r) { |
1225 | DMERR("could not register integrity profile."); | |
1226 | return r; | |
1227 | } | |
1228 | ||
17e149b8 | 1229 | r = dm_table_alloc_md_mempools(t, t->md); |
26803b9f WD |
1230 | if (r) |
1231 | DMERR("unable to allocate mempools"); | |
1232 | ||
1233 | return r; | |
1234 | } | |
1235 | ||
48c9c27b | 1236 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1237 | void dm_table_event_callback(struct dm_table *t, |
1238 | void (*fn)(void *), void *context) | |
1239 | { | |
48c9c27b | 1240 | mutex_lock(&_event_lock); |
1da177e4 LT |
1241 | t->event_fn = fn; |
1242 | t->event_context = context; | |
48c9c27b | 1243 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1244 | } |
1245 | ||
1246 | void dm_table_event(struct dm_table *t) | |
1247 | { | |
1248 | /* | |
1249 | * You can no longer call dm_table_event() from interrupt | |
1250 | * context, use a bottom half instead. | |
1251 | */ | |
1252 | BUG_ON(in_interrupt()); | |
1253 | ||
48c9c27b | 1254 | mutex_lock(&_event_lock); |
1da177e4 LT |
1255 | if (t->event_fn) |
1256 | t->event_fn(t->event_context); | |
48c9c27b | 1257 | mutex_unlock(&_event_lock); |
1da177e4 | 1258 | } |
08649012 | 1259 | EXPORT_SYMBOL(dm_table_event); |
1da177e4 LT |
1260 | |
1261 | sector_t dm_table_get_size(struct dm_table *t) | |
1262 | { | |
1263 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1264 | } | |
08649012 | 1265 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 LT |
1266 | |
1267 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1268 | { | |
14353539 | 1269 | if (index >= t->num_targets) |
1da177e4 LT |
1270 | return NULL; |
1271 | ||
1272 | return t->targets + index; | |
1273 | } | |
1274 | ||
1275 | /* | |
1276 | * Search the btree for the correct target. | |
512875bd JN |
1277 | * |
1278 | * Caller should check returned pointer with dm_target_is_valid() | |
1279 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1280 | */ |
1281 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1282 | { | |
1283 | unsigned int l, n = 0, k = 0; | |
1284 | sector_t *node; | |
1285 | ||
1286 | for (l = 0; l < t->depth; l++) { | |
1287 | n = get_child(n, k); | |
1288 | node = get_node(t, l, n); | |
1289 | ||
1290 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1291 | if (node[k] >= sector) | |
1292 | break; | |
1293 | } | |
1294 | ||
1295 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1296 | } | |
1297 | ||
3ae70656 MS |
1298 | static int count_device(struct dm_target *ti, struct dm_dev *dev, |
1299 | sector_t start, sector_t len, void *data) | |
1300 | { | |
1301 | unsigned *num_devices = data; | |
1302 | ||
1303 | (*num_devices)++; | |
1304 | ||
1305 | return 0; | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * Check whether a table has no data devices attached using each | |
1310 | * target's iterate_devices method. | |
1311 | * Returns false if the result is unknown because a target doesn't | |
1312 | * support iterate_devices. | |
1313 | */ | |
1314 | bool dm_table_has_no_data_devices(struct dm_table *table) | |
1315 | { | |
1316 | struct dm_target *uninitialized_var(ti); | |
1317 | unsigned i = 0, num_devices = 0; | |
1318 | ||
1319 | while (i < dm_table_get_num_targets(table)) { | |
1320 | ti = dm_table_get_target(table, i++); | |
1321 | ||
1322 | if (!ti->type->iterate_devices) | |
1323 | return false; | |
1324 | ||
1325 | ti->type->iterate_devices(ti, count_device, &num_devices); | |
1326 | if (num_devices) | |
1327 | return false; | |
1328 | } | |
1329 | ||
1330 | return true; | |
1331 | } | |
1332 | ||
754c5fc7 MS |
1333 | /* |
1334 | * Establish the new table's queue_limits and validate them. | |
1335 | */ | |
1336 | int dm_calculate_queue_limits(struct dm_table *table, | |
1337 | struct queue_limits *limits) | |
1338 | { | |
1339 | struct dm_target *uninitialized_var(ti); | |
1340 | struct queue_limits ti_limits; | |
1341 | unsigned i = 0; | |
1342 | ||
b1bd055d | 1343 | blk_set_stacking_limits(limits); |
754c5fc7 MS |
1344 | |
1345 | while (i < dm_table_get_num_targets(table)) { | |
b1bd055d | 1346 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
1347 | |
1348 | ti = dm_table_get_target(table, i++); | |
1349 | ||
1350 | if (!ti->type->iterate_devices) | |
1351 | goto combine_limits; | |
1352 | ||
1353 | /* | |
1354 | * Combine queue limits of all the devices this target uses. | |
1355 | */ | |
1356 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1357 | &ti_limits); | |
1358 | ||
40bea431 MS |
1359 | /* Set I/O hints portion of queue limits */ |
1360 | if (ti->type->io_hints) | |
1361 | ti->type->io_hints(ti, &ti_limits); | |
1362 | ||
754c5fc7 MS |
1363 | /* |
1364 | * Check each device area is consistent with the target's | |
1365 | * overall queue limits. | |
1366 | */ | |
f6a1ed10 MP |
1367 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1368 | &ti_limits)) | |
754c5fc7 MS |
1369 | return -EINVAL; |
1370 | ||
1371 | combine_limits: | |
1372 | /* | |
1373 | * Merge this target's queue limits into the overall limits | |
1374 | * for the table. | |
1375 | */ | |
1376 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1377 | DMWARN("%s: adding target device " |
754c5fc7 | 1378 | "(start sect %llu len %llu) " |
b27d7f16 | 1379 | "caused an alignment inconsistency", |
754c5fc7 MS |
1380 | dm_device_name(table->md), |
1381 | (unsigned long long) ti->begin, | |
1382 | (unsigned long long) ti->len); | |
1383 | } | |
1384 | ||
1385 | return validate_hardware_logical_block_alignment(table, limits); | |
1386 | } | |
1387 | ||
9c47008d | 1388 | /* |
25520d55 MP |
1389 | * Verify that all devices have an integrity profile that matches the |
1390 | * DM device's registered integrity profile. If the profiles don't | |
1391 | * match then unregister the DM device's integrity profile. | |
9c47008d | 1392 | */ |
25520d55 | 1393 | static void dm_table_verify_integrity(struct dm_table *t) |
9c47008d | 1394 | { |
a63a5cf8 | 1395 | struct gendisk *template_disk = NULL; |
9c47008d | 1396 | |
25520d55 MP |
1397 | if (t->integrity_supported) { |
1398 | /* | |
1399 | * Verify that the original integrity profile | |
1400 | * matches all the devices in this table. | |
1401 | */ | |
1402 | template_disk = dm_table_get_integrity_disk(t); | |
1403 | if (template_disk && | |
1404 | blk_integrity_compare(dm_disk(t->md), template_disk) >= 0) | |
1405 | return; | |
1406 | } | |
9c47008d | 1407 | |
25520d55 | 1408 | if (integrity_profile_exists(dm_disk(t->md))) { |
876fbba1 MS |
1409 | DMWARN("%s: unable to establish an integrity profile", |
1410 | dm_device_name(t->md)); | |
25520d55 MP |
1411 | blk_integrity_unregister(dm_disk(t->md)); |
1412 | } | |
9c47008d MP |
1413 | } |
1414 | ||
ed8b752b MS |
1415 | static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, |
1416 | sector_t start, sector_t len, void *data) | |
1417 | { | |
c888a8f9 | 1418 | unsigned long flush = (unsigned long) data; |
ed8b752b MS |
1419 | struct request_queue *q = bdev_get_queue(dev->bdev); |
1420 | ||
c888a8f9 | 1421 | return q && (q->queue_flags & flush); |
ed8b752b MS |
1422 | } |
1423 | ||
c888a8f9 | 1424 | static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush) |
ed8b752b MS |
1425 | { |
1426 | struct dm_target *ti; | |
1427 | unsigned i = 0; | |
1428 | ||
1429 | /* | |
1430 | * Require at least one underlying device to support flushes. | |
1431 | * t->devices includes internal dm devices such as mirror logs | |
1432 | * so we need to use iterate_devices here, which targets | |
1433 | * supporting flushes must provide. | |
1434 | */ | |
1435 | while (i < dm_table_get_num_targets(t)) { | |
1436 | ti = dm_table_get_target(t, i++); | |
1437 | ||
55a62eef | 1438 | if (!ti->num_flush_bios) |
ed8b752b MS |
1439 | continue; |
1440 | ||
0e9c24ed | 1441 | if (ti->flush_supported) |
7f61f5a0 | 1442 | return true; |
0e9c24ed | 1443 | |
ed8b752b | 1444 | if (ti->type->iterate_devices && |
c888a8f9 | 1445 | ti->type->iterate_devices(ti, device_flush_capable, (void *) flush)) |
7f61f5a0 | 1446 | return true; |
ed8b752b MS |
1447 | } |
1448 | ||
7f61f5a0 | 1449 | return false; |
ed8b752b MS |
1450 | } |
1451 | ||
983c7db3 MB |
1452 | static bool dm_table_discard_zeroes_data(struct dm_table *t) |
1453 | { | |
1454 | struct dm_target *ti; | |
1455 | unsigned i = 0; | |
1456 | ||
1457 | /* Ensure that all targets supports discard_zeroes_data. */ | |
1458 | while (i < dm_table_get_num_targets(t)) { | |
1459 | ti = dm_table_get_target(t, i++); | |
1460 | ||
1461 | if (ti->discard_zeroes_data_unsupported) | |
7f61f5a0 | 1462 | return false; |
983c7db3 MB |
1463 | } |
1464 | ||
7f61f5a0 | 1465 | return true; |
983c7db3 MB |
1466 | } |
1467 | ||
4693c966 MSB |
1468 | static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev, |
1469 | sector_t start, sector_t len, void *data) | |
1470 | { | |
1471 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1472 | ||
1473 | return q && blk_queue_nonrot(q); | |
1474 | } | |
1475 | ||
c3c4555e MB |
1476 | static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, |
1477 | sector_t start, sector_t len, void *data) | |
1478 | { | |
1479 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1480 | ||
1481 | return q && !blk_queue_add_random(q); | |
1482 | } | |
1483 | ||
200612ec JM |
1484 | static int queue_supports_sg_merge(struct dm_target *ti, struct dm_dev *dev, |
1485 | sector_t start, sector_t len, void *data) | |
1486 | { | |
1487 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1488 | ||
1489 | return q && !test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags); | |
1490 | } | |
1491 | ||
c3c4555e MB |
1492 | static bool dm_table_all_devices_attribute(struct dm_table *t, |
1493 | iterate_devices_callout_fn func) | |
4693c966 MSB |
1494 | { |
1495 | struct dm_target *ti; | |
1496 | unsigned i = 0; | |
1497 | ||
4693c966 MSB |
1498 | while (i < dm_table_get_num_targets(t)) { |
1499 | ti = dm_table_get_target(t, i++); | |
1500 | ||
1501 | if (!ti->type->iterate_devices || | |
c3c4555e | 1502 | !ti->type->iterate_devices(ti, func, NULL)) |
7f61f5a0 | 1503 | return false; |
4693c966 MSB |
1504 | } |
1505 | ||
7f61f5a0 | 1506 | return true; |
4693c966 MSB |
1507 | } |
1508 | ||
d54eaa5a MS |
1509 | static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev, |
1510 | sector_t start, sector_t len, void *data) | |
1511 | { | |
1512 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1513 | ||
1514 | return q && !q->limits.max_write_same_sectors; | |
1515 | } | |
1516 | ||
1517 | static bool dm_table_supports_write_same(struct dm_table *t) | |
1518 | { | |
1519 | struct dm_target *ti; | |
1520 | unsigned i = 0; | |
1521 | ||
1522 | while (i < dm_table_get_num_targets(t)) { | |
1523 | ti = dm_table_get_target(t, i++); | |
1524 | ||
55a62eef | 1525 | if (!ti->num_write_same_bios) |
d54eaa5a MS |
1526 | return false; |
1527 | ||
1528 | if (!ti->type->iterate_devices || | |
dc019b21 | 1529 | ti->type->iterate_devices(ti, device_not_write_same_capable, NULL)) |
d54eaa5a MS |
1530 | return false; |
1531 | } | |
1532 | ||
1533 | return true; | |
1534 | } | |
1535 | ||
ac62d620 CH |
1536 | static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev, |
1537 | sector_t start, sector_t len, void *data) | |
1538 | { | |
1539 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1540 | ||
1541 | return q && !q->limits.max_write_zeroes_sectors; | |
1542 | } | |
1543 | ||
1544 | static bool dm_table_supports_write_zeroes(struct dm_table *t) | |
1545 | { | |
1546 | struct dm_target *ti; | |
1547 | unsigned i = 0; | |
1548 | ||
1549 | while (i < dm_table_get_num_targets(t)) { | |
1550 | ti = dm_table_get_target(t, i++); | |
1551 | ||
1552 | if (!ti->num_write_zeroes_bios) | |
1553 | return false; | |
1554 | ||
1555 | if (!ti->type->iterate_devices || | |
1556 | ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL)) | |
1557 | return false; | |
1558 | } | |
1559 | ||
1560 | return true; | |
1561 | } | |
1562 | ||
1563 | ||
a7ffb6a5 MP |
1564 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1565 | sector_t start, sector_t len, void *data) | |
1566 | { | |
1567 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1568 | ||
1569 | return q && blk_queue_discard(q); | |
1570 | } | |
1571 | ||
1572 | static bool dm_table_supports_discards(struct dm_table *t) | |
1573 | { | |
1574 | struct dm_target *ti; | |
1575 | unsigned i = 0; | |
1576 | ||
1577 | /* | |
1578 | * Unless any target used by the table set discards_supported, | |
1579 | * require at least one underlying device to support discards. | |
1580 | * t->devices includes internal dm devices such as mirror logs | |
1581 | * so we need to use iterate_devices here, which targets | |
1582 | * supporting discard selectively must provide. | |
1583 | */ | |
1584 | while (i < dm_table_get_num_targets(t)) { | |
1585 | ti = dm_table_get_target(t, i++); | |
1586 | ||
1587 | if (!ti->num_discard_bios) | |
1588 | continue; | |
1589 | ||
1590 | if (ti->discards_supported) | |
7f61f5a0 | 1591 | return true; |
a7ffb6a5 MP |
1592 | |
1593 | if (ti->type->iterate_devices && | |
1594 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
7f61f5a0 | 1595 | return true; |
a7ffb6a5 MP |
1596 | } |
1597 | ||
7f61f5a0 | 1598 | return false; |
a7ffb6a5 MP |
1599 | } |
1600 | ||
754c5fc7 MS |
1601 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1602 | struct queue_limits *limits) | |
1da177e4 | 1603 | { |
519a7e16 | 1604 | bool wc = false, fua = false; |
ed8b752b | 1605 | |
1da177e4 | 1606 | /* |
1197764e | 1607 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1608 | */ |
754c5fc7 | 1609 | q->limits = *limits; |
c9a3f6d6 | 1610 | |
5ae89a87 MS |
1611 | if (!dm_table_supports_discards(t)) |
1612 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1613 | else | |
1614 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1615 | ||
c888a8f9 | 1616 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) { |
519a7e16 | 1617 | wc = true; |
c888a8f9 | 1618 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA))) |
519a7e16 | 1619 | fua = true; |
ed8b752b | 1620 | } |
519a7e16 | 1621 | blk_queue_write_cache(q, wc, fua); |
ed8b752b | 1622 | |
983c7db3 MB |
1623 | if (!dm_table_discard_zeroes_data(t)) |
1624 | q->limits.discard_zeroes_data = 0; | |
1625 | ||
c3c4555e MB |
1626 | /* Ensure that all underlying devices are non-rotational. */ |
1627 | if (dm_table_all_devices_attribute(t, device_is_nonrot)) | |
4693c966 MSB |
1628 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q); |
1629 | else | |
1630 | queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q); | |
1631 | ||
d54eaa5a MS |
1632 | if (!dm_table_supports_write_same(t)) |
1633 | q->limits.max_write_same_sectors = 0; | |
ac62d620 CH |
1634 | if (!dm_table_supports_write_zeroes(t)) |
1635 | q->limits.max_write_zeroes_sectors = 0; | |
c1a94672 | 1636 | |
200612ec JM |
1637 | if (dm_table_all_devices_attribute(t, queue_supports_sg_merge)) |
1638 | queue_flag_clear_unlocked(QUEUE_FLAG_NO_SG_MERGE, q); | |
1639 | else | |
1640 | queue_flag_set_unlocked(QUEUE_FLAG_NO_SG_MERGE, q); | |
1641 | ||
25520d55 | 1642 | dm_table_verify_integrity(t); |
e6ee8c0b | 1643 | |
c3c4555e MB |
1644 | /* |
1645 | * Determine whether or not this queue's I/O timings contribute | |
1646 | * to the entropy pool, Only request-based targets use this. | |
1647 | * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not | |
1648 | * have it set. | |
1649 | */ | |
1650 | if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random)) | |
1651 | queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, q); | |
1652 | ||
e6ee8c0b KU |
1653 | /* |
1654 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1655 | * visible to other CPUs because, once the flag is set, incoming bios | |
1656 | * are processed by request-based dm, which refers to the queue | |
1657 | * settings. | |
1658 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1659 | * md->deferred where queue settings are not needed yet. | |
1660 | * Those bios are passed to request-based dm at the resume time. | |
1661 | */ | |
1662 | smp_mb(); | |
1663 | if (dm_table_request_based(t)) | |
1664 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1665 | } |
1666 | ||
1667 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1668 | { | |
1669 | return t->num_targets; | |
1670 | } | |
1671 | ||
1672 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1673 | { | |
1674 | return &t->devices; | |
1675 | } | |
1676 | ||
aeb5d727 | 1677 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1678 | { |
1679 | return t->mode; | |
1680 | } | |
08649012 | 1681 | EXPORT_SYMBOL(dm_table_get_mode); |
1da177e4 | 1682 | |
d67ee213 MS |
1683 | enum suspend_mode { |
1684 | PRESUSPEND, | |
1685 | PRESUSPEND_UNDO, | |
1686 | POSTSUSPEND, | |
1687 | }; | |
1688 | ||
1689 | static void suspend_targets(struct dm_table *t, enum suspend_mode mode) | |
1da177e4 LT |
1690 | { |
1691 | int i = t->num_targets; | |
1692 | struct dm_target *ti = t->targets; | |
1693 | ||
1694 | while (i--) { | |
d67ee213 MS |
1695 | switch (mode) { |
1696 | case PRESUSPEND: | |
1697 | if (ti->type->presuspend) | |
1698 | ti->type->presuspend(ti); | |
1699 | break; | |
1700 | case PRESUSPEND_UNDO: | |
1701 | if (ti->type->presuspend_undo) | |
1702 | ti->type->presuspend_undo(ti); | |
1703 | break; | |
1704 | case POSTSUSPEND: | |
1da177e4 LT |
1705 | if (ti->type->postsuspend) |
1706 | ti->type->postsuspend(ti); | |
d67ee213 MS |
1707 | break; |
1708 | } | |
1da177e4 LT |
1709 | ti++; |
1710 | } | |
1711 | } | |
1712 | ||
1713 | void dm_table_presuspend_targets(struct dm_table *t) | |
1714 | { | |
cf222b37 AK |
1715 | if (!t) |
1716 | return; | |
1717 | ||
d67ee213 MS |
1718 | suspend_targets(t, PRESUSPEND); |
1719 | } | |
1720 | ||
1721 | void dm_table_presuspend_undo_targets(struct dm_table *t) | |
1722 | { | |
1723 | if (!t) | |
1724 | return; | |
1725 | ||
1726 | suspend_targets(t, PRESUSPEND_UNDO); | |
1da177e4 LT |
1727 | } |
1728 | ||
1729 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1730 | { | |
cf222b37 AK |
1731 | if (!t) |
1732 | return; | |
1733 | ||
d67ee213 | 1734 | suspend_targets(t, POSTSUSPEND); |
1da177e4 LT |
1735 | } |
1736 | ||
8757b776 | 1737 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1738 | { |
8757b776 MB |
1739 | int i, r = 0; |
1740 | ||
1741 | for (i = 0; i < t->num_targets; i++) { | |
1742 | struct dm_target *ti = t->targets + i; | |
1743 | ||
1744 | if (!ti->type->preresume) | |
1745 | continue; | |
1746 | ||
1747 | r = ti->type->preresume(ti); | |
7833b08e MS |
1748 | if (r) { |
1749 | DMERR("%s: %s: preresume failed, error = %d", | |
1750 | dm_device_name(t->md), ti->type->name, r); | |
8757b776 | 1751 | return r; |
7833b08e | 1752 | } |
8757b776 | 1753 | } |
1da177e4 LT |
1754 | |
1755 | for (i = 0; i < t->num_targets; i++) { | |
1756 | struct dm_target *ti = t->targets + i; | |
1757 | ||
1758 | if (ti->type->resume) | |
1759 | ti->type->resume(ti); | |
1760 | } | |
8757b776 MB |
1761 | |
1762 | return 0; | |
1da177e4 LT |
1763 | } |
1764 | ||
9d357b07 N |
1765 | void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb) |
1766 | { | |
1767 | list_add(&cb->list, &t->target_callbacks); | |
1768 | } | |
1769 | EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks); | |
1770 | ||
1da177e4 LT |
1771 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) |
1772 | { | |
82b1519b | 1773 | struct dm_dev_internal *dd; |
afb24528 | 1774 | struct list_head *devices = dm_table_get_devices(t); |
9d357b07 | 1775 | struct dm_target_callbacks *cb; |
1da177e4 LT |
1776 | int r = 0; |
1777 | ||
afb24528 | 1778 | list_for_each_entry(dd, devices, list) { |
86f1152b | 1779 | struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev); |
0c2322e4 AK |
1780 | char b[BDEVNAME_SIZE]; |
1781 | ||
1782 | if (likely(q)) | |
dc3b17cc | 1783 | r |= bdi_congested(q->backing_dev_info, bdi_bits); |
0c2322e4 AK |
1784 | else |
1785 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1786 | dm_device_name(t->md), | |
86f1152b | 1787 | bdevname(dd->dm_dev->bdev, b)); |
1da177e4 LT |
1788 | } |
1789 | ||
9d357b07 N |
1790 | list_for_each_entry(cb, &t->target_callbacks, list) |
1791 | if (cb->congested_fn) | |
1792 | r |= cb->congested_fn(cb, bdi_bits); | |
1793 | ||
1da177e4 LT |
1794 | return r; |
1795 | } | |
1796 | ||
1134e5ae MA |
1797 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1798 | { | |
1134e5ae MA |
1799 | return t->md; |
1800 | } | |
08649012 | 1801 | EXPORT_SYMBOL(dm_table_get_md); |
1134e5ae | 1802 | |
9974fa2c MS |
1803 | void dm_table_run_md_queue_async(struct dm_table *t) |
1804 | { | |
1805 | struct mapped_device *md; | |
1806 | struct request_queue *queue; | |
1807 | unsigned long flags; | |
1808 | ||
1809 | if (!dm_table_request_based(t)) | |
1810 | return; | |
1811 | ||
1812 | md = dm_table_get_md(t); | |
1813 | queue = dm_get_md_queue(md); | |
1814 | if (queue) { | |
bfebd1cd MS |
1815 | if (queue->mq_ops) |
1816 | blk_mq_run_hw_queues(queue, true); | |
1817 | else { | |
1818 | spin_lock_irqsave(queue->queue_lock, flags); | |
1819 | blk_run_queue_async(queue); | |
1820 | spin_unlock_irqrestore(queue->queue_lock, flags); | |
1821 | } | |
9974fa2c MS |
1822 | } |
1823 | } | |
1824 | EXPORT_SYMBOL(dm_table_run_md_queue_async); | |
1825 |