Commit | Line | Data |
---|---|---|
3bd94003 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * Copyright (C) 2001 Sistina Software (UK) Limited. | |
d5816876 | 4 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
1da177e4 LT |
5 | * |
6 | * This file is released under the GPL. | |
7 | */ | |
8 | ||
4cc96131 | 9 | #include "dm-core.h" |
e810cb78 | 10 | #include "dm-rq.h" |
1da177e4 LT |
11 | |
12 | #include <linux/module.h> | |
13 | #include <linux/vmalloc.h> | |
14 | #include <linux/blkdev.h> | |
fe45e630 | 15 | #include <linux/blk-integrity.h> |
1da177e4 LT |
16 | #include <linux/namei.h> |
17 | #include <linux/ctype.h> | |
e7d2860b | 18 | #include <linux/string.h> |
1da177e4 LT |
19 | #include <linux/slab.h> |
20 | #include <linux/interrupt.h> | |
48c9c27b | 21 | #include <linux/mutex.h> |
d5816876 | 22 | #include <linux/delay.h> |
60063497 | 23 | #include <linux/atomic.h> |
bfebd1cd | 24 | #include <linux/blk-mq.h> |
644bda6f | 25 | #include <linux/mount.h> |
273752c9 | 26 | #include <linux/dax.h> |
1da177e4 | 27 | |
72d94861 AK |
28 | #define DM_MSG_PREFIX "table" |
29 | ||
1da177e4 LT |
30 | #define NODE_SIZE L1_CACHE_BYTES |
31 | #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t)) | |
32 | #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1) | |
33 | ||
1da177e4 LT |
34 | /* |
35 | * Similar to ceiling(log_size(n)) | |
36 | */ | |
37 | static unsigned int int_log(unsigned int n, unsigned int base) | |
38 | { | |
39 | int result = 0; | |
40 | ||
41 | while (n > 1) { | |
42 | n = dm_div_up(n, base); | |
43 | result++; | |
44 | } | |
45 | ||
46 | return result; | |
47 | } | |
48 | ||
1da177e4 LT |
49 | /* |
50 | * Calculate the index of the child node of the n'th node k'th key. | |
51 | */ | |
52 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
53 | { | |
54 | return (n * CHILDREN_PER_NODE) + k; | |
55 | } | |
56 | ||
57 | /* | |
58 | * Return the n'th node of level l from table t. | |
59 | */ | |
60 | static inline sector_t *get_node(struct dm_table *t, | |
61 | unsigned int l, unsigned int n) | |
62 | { | |
63 | return t->index[l] + (n * KEYS_PER_NODE); | |
64 | } | |
65 | ||
66 | /* | |
67 | * Return the highest key that you could lookup from the n'th | |
68 | * node on level l of the btree. | |
69 | */ | |
70 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
71 | { | |
72 | for (; l < t->depth - 1; l++) | |
73 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
74 | ||
75 | if (n >= t->counts[l]) | |
255e2646 | 76 | return (sector_t) -1; |
1da177e4 LT |
77 | |
78 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
79 | } | |
80 | ||
81 | /* | |
82 | * Fills in a level of the btree based on the highs of the level | |
83 | * below it. | |
84 | */ | |
85 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
86 | { | |
87 | unsigned int n, k; | |
88 | sector_t *node; | |
89 | ||
90 | for (n = 0U; n < t->counts[l]; n++) { | |
91 | node = get_node(t, l, n); | |
92 | ||
93 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
94 | node[k] = high(t, l + 1, get_child(n, k)); | |
95 | } | |
96 | ||
97 | return 0; | |
98 | } | |
99 | ||
1da177e4 LT |
100 | /* |
101 | * highs, and targets are managed as dynamic arrays during a | |
102 | * table load. | |
103 | */ | |
104 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
105 | { | |
106 | sector_t *n_highs; | |
107 | struct dm_target *n_targets; | |
1da177e4 LT |
108 | |
109 | /* | |
110 | * Allocate both the target array and offset array at once. | |
111 | */ | |
7a35693a MWO |
112 | n_highs = kvcalloc(num, sizeof(struct dm_target) + sizeof(sector_t), |
113 | GFP_KERNEL); | |
1da177e4 LT |
114 | if (!n_highs) |
115 | return -ENOMEM; | |
116 | ||
117 | n_targets = (struct dm_target *) (n_highs + num); | |
118 | ||
57a2f238 | 119 | memset(n_highs, -1, sizeof(*n_highs) * num); |
7a35693a | 120 | kvfree(t->highs); |
1da177e4 LT |
121 | |
122 | t->num_allocated = num; | |
123 | t->highs = n_highs; | |
124 | t->targets = n_targets; | |
125 | ||
126 | return 0; | |
127 | } | |
128 | ||
05bdb996 | 129 | int dm_table_create(struct dm_table **result, blk_mode_t mode, |
86a3238c | 130 | unsigned int num_targets, struct mapped_device *md) |
1da177e4 | 131 | { |
094262db | 132 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
133 | |
134 | if (!t) | |
135 | return -ENOMEM; | |
136 | ||
1da177e4 | 137 | INIT_LIST_HEAD(&t->devices); |
f6007dce | 138 | init_rwsem(&t->devices_lock); |
1da177e4 LT |
139 | |
140 | if (!num_targets) | |
141 | num_targets = KEYS_PER_NODE; | |
142 | ||
143 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
144 | ||
5b2d0657 MP |
145 | if (!num_targets) { |
146 | kfree(t); | |
147 | return -ENOMEM; | |
148 | } | |
149 | ||
1da177e4 LT |
150 | if (alloc_targets(t, num_targets)) { |
151 | kfree(t); | |
1da177e4 LT |
152 | return -ENOMEM; |
153 | } | |
154 | ||
e83068a5 | 155 | t->type = DM_TYPE_NONE; |
1da177e4 | 156 | t->mode = mode; |
1134e5ae | 157 | t->md = md; |
1da177e4 LT |
158 | *result = t; |
159 | return 0; | |
160 | } | |
161 | ||
86f1152b | 162 | static void free_devices(struct list_head *devices, struct mapped_device *md) |
1da177e4 LT |
163 | { |
164 | struct list_head *tmp, *next; | |
165 | ||
afb24528 | 166 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
167 | struct dm_dev_internal *dd = |
168 | list_entry(tmp, struct dm_dev_internal, list); | |
86f1152b BM |
169 | DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s", |
170 | dm_device_name(md), dd->dm_dev->name); | |
171 | dm_put_table_device(md, dd->dm_dev); | |
1da177e4 LT |
172 | kfree(dd); |
173 | } | |
174 | } | |
175 | ||
cb77cb5a | 176 | static void dm_table_destroy_crypto_profile(struct dm_table *t); |
aa6ce87a | 177 | |
d5816876 | 178 | void dm_table_destroy(struct dm_table *t) |
1da177e4 | 179 | { |
a7940155 AK |
180 | if (!t) |
181 | return; | |
182 | ||
26803b9f | 183 | /* free the indexes */ |
1da177e4 | 184 | if (t->depth >= 2) |
7a35693a | 185 | kvfree(t->index[t->depth - 2]); |
1da177e4 LT |
186 | |
187 | /* free the targets */ | |
564b5c54 MS |
188 | for (unsigned int i = 0; i < t->num_targets; i++) { |
189 | struct dm_target *ti = dm_table_get_target(t, i); | |
1da177e4 | 190 | |
564b5c54 MS |
191 | if (ti->type->dtr) |
192 | ti->type->dtr(ti); | |
1da177e4 | 193 | |
564b5c54 | 194 | dm_put_target_type(ti->type); |
1da177e4 LT |
195 | } |
196 | ||
7a35693a | 197 | kvfree(t->highs); |
1da177e4 LT |
198 | |
199 | /* free the device list */ | |
86f1152b | 200 | free_devices(&t->devices, t->md); |
1da177e4 | 201 | |
e6ee8c0b KU |
202 | dm_free_md_mempools(t->mempools); |
203 | ||
cb77cb5a | 204 | dm_table_destroy_crypto_profile(t); |
aa6ce87a | 205 | |
1da177e4 LT |
206 | kfree(t); |
207 | } | |
208 | ||
1da177e4 LT |
209 | /* |
210 | * See if we've already got a device in the list. | |
211 | */ | |
82b1519b | 212 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 213 | { |
82b1519b | 214 | struct dm_dev_internal *dd; |
1da177e4 | 215 | |
43be9c74 | 216 | list_for_each_entry(dd, l, list) |
86f1152b | 217 | if (dd->dm_dev->bdev->bd_dev == dev) |
1da177e4 LT |
218 | return dd; |
219 | ||
220 | return NULL; | |
221 | } | |
222 | ||
1da177e4 | 223 | /* |
f6a1ed10 | 224 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 225 | */ |
f6a1ed10 MP |
226 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
227 | sector_t start, sector_t len, void *data) | |
1da177e4 | 228 | { |
754c5fc7 MS |
229 | struct queue_limits *limits = data; |
230 | struct block_device *bdev = dev->bdev; | |
6dcbb52c | 231 | sector_t dev_size = bdev_nr_sectors(bdev); |
02acc3a4 | 232 | unsigned short logical_block_size_sectors = |
754c5fc7 | 233 | limits->logical_block_size >> SECTOR_SHIFT; |
2cd54d9b MA |
234 | |
235 | if (!dev_size) | |
f6a1ed10 | 236 | return 0; |
2cd54d9b | 237 | |
5dea271b | 238 | if ((start >= dev_size) || (start + len > dev_size)) { |
2e84fecf | 239 | DMERR("%s: %pg too small for target: start=%llu, len=%llu, dev_size=%llu", |
43e6c111 MP |
240 | dm_device_name(ti->table->md), bdev, |
241 | (unsigned long long)start, | |
242 | (unsigned long long)len, | |
243 | (unsigned long long)dev_size); | |
f6a1ed10 | 244 | return 1; |
02acc3a4 MS |
245 | } |
246 | ||
dd88d313 DLM |
247 | /* |
248 | * If the target is mapped to zoned block device(s), check | |
249 | * that the zones are not partially mapped. | |
250 | */ | |
dd73c320 | 251 | if (bdev_is_zoned(bdev)) { |
dd88d313 DLM |
252 | unsigned int zone_sectors = bdev_zone_sectors(bdev); |
253 | ||
254 | if (start & (zone_sectors - 1)) { | |
43e6c111 MP |
255 | DMERR("%s: start=%llu not aligned to h/w zone size %u of %pg", |
256 | dm_device_name(ti->table->md), | |
257 | (unsigned long long)start, | |
258 | zone_sectors, bdev); | |
dd88d313 DLM |
259 | return 1; |
260 | } | |
261 | ||
262 | /* | |
263 | * Note: The last zone of a zoned block device may be smaller | |
264 | * than other zones. So for a target mapping the end of a | |
265 | * zoned block device with such a zone, len would not be zone | |
266 | * aligned. We do not allow such last smaller zone to be part | |
267 | * of the mapping here to ensure that mappings with multiple | |
268 | * devices do not end up with a smaller zone in the middle of | |
269 | * the sector range. | |
270 | */ | |
271 | if (len & (zone_sectors - 1)) { | |
43e6c111 MP |
272 | DMERR("%s: len=%llu not aligned to h/w zone size %u of %pg", |
273 | dm_device_name(ti->table->md), | |
274 | (unsigned long long)len, | |
275 | zone_sectors, bdev); | |
dd88d313 DLM |
276 | return 1; |
277 | } | |
278 | } | |
279 | ||
02acc3a4 | 280 | if (logical_block_size_sectors <= 1) |
f6a1ed10 | 281 | return 0; |
02acc3a4 MS |
282 | |
283 | if (start & (logical_block_size_sectors - 1)) { | |
2e84fecf | 284 | DMERR("%s: start=%llu not aligned to h/w logical block size %u of %pg", |
43e6c111 MP |
285 | dm_device_name(ti->table->md), |
286 | (unsigned long long)start, | |
287 | limits->logical_block_size, bdev); | |
f6a1ed10 | 288 | return 1; |
02acc3a4 MS |
289 | } |
290 | ||
5dea271b | 291 | if (len & (logical_block_size_sectors - 1)) { |
2e84fecf | 292 | DMERR("%s: len=%llu not aligned to h/w logical block size %u of %pg", |
43e6c111 MP |
293 | dm_device_name(ti->table->md), |
294 | (unsigned long long)len, | |
295 | limits->logical_block_size, bdev); | |
f6a1ed10 | 296 | return 1; |
02acc3a4 MS |
297 | } |
298 | ||
f6a1ed10 | 299 | return 0; |
1da177e4 LT |
300 | } |
301 | ||
302 | /* | |
570b9d96 | 303 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 304 | * careful to leave things as they were if we fail to reopen the |
570b9d96 | 305 | * device and not to touch the existing bdev field in case |
21cf8661 | 306 | * it is accessed concurrently. |
1da177e4 | 307 | */ |
05bdb996 | 308 | static int upgrade_mode(struct dm_dev_internal *dd, blk_mode_t new_mode, |
82b1519b | 309 | struct mapped_device *md) |
1da177e4 LT |
310 | { |
311 | int r; | |
86f1152b | 312 | struct dm_dev *old_dev, *new_dev; |
1da177e4 | 313 | |
86f1152b | 314 | old_dev = dd->dm_dev; |
570b9d96 | 315 | |
86f1152b BM |
316 | r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, |
317 | dd->dm_dev->mode | new_mode, &new_dev); | |
570b9d96 AK |
318 | if (r) |
319 | return r; | |
1da177e4 | 320 | |
86f1152b BM |
321 | dd->dm_dev = new_dev; |
322 | dm_put_table_device(md, old_dev); | |
1da177e4 | 323 | |
570b9d96 | 324 | return 0; |
1da177e4 LT |
325 | } |
326 | ||
327 | /* | |
328 | * Add a device to the list, or just increment the usage count if | |
329 | * it's already present. | |
7a126d5b CH |
330 | * |
331 | * Note: the __ref annotation is because this function can call the __init | |
332 | * marked early_lookup_bdev when called during early boot code from dm-init.c. | |
1da177e4 | 333 | */ |
05bdb996 | 334 | int __ref dm_get_device(struct dm_target *ti, const char *path, blk_mode_t mode, |
08649012 | 335 | struct dm_dev **result) |
1da177e4 LT |
336 | { |
337 | int r; | |
4df2bf46 | 338 | dev_t dev; |
809b1e49 HR |
339 | unsigned int major, minor; |
340 | char dummy; | |
82b1519b | 341 | struct dm_dev_internal *dd; |
08649012 | 342 | struct dm_table *t = ti->table; |
1da177e4 | 343 | |
547bc926 | 344 | BUG_ON(!t); |
1da177e4 | 345 | |
809b1e49 HR |
346 | if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) { |
347 | /* Extract the major/minor numbers */ | |
348 | dev = MKDEV(major, minor); | |
349 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
350 | return -EOVERFLOW; | |
351 | } else { | |
d4a28d7d | 352 | r = lookup_bdev(path, &dev); |
7a126d5b CH |
353 | #ifndef MODULE |
354 | if (r && system_state < SYSTEM_RUNNING) | |
d4a28d7d | 355 | r = early_lookup_bdev(path, &dev); |
7a126d5b | 356 | #endif |
d4a28d7d CH |
357 | if (r) |
358 | return r; | |
809b1e49 | 359 | } |
d1c0e158 BM |
360 | if (dev == disk_devt(t->md->disk)) |
361 | return -EINVAL; | |
1da177e4 | 362 | |
f6007dce MP |
363 | down_write(&t->devices_lock); |
364 | ||
1da177e4 LT |
365 | dd = find_device(&t->devices, dev); |
366 | if (!dd) { | |
367 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
f6007dce MP |
368 | if (!dd) { |
369 | r = -ENOMEM; | |
370 | goto unlock_ret_r; | |
371 | } | |
1da177e4 | 372 | |
d715fa23 HM |
373 | r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev); |
374 | if (r) { | |
1da177e4 | 375 | kfree(dd); |
f6007dce | 376 | goto unlock_ret_r; |
1da177e4 LT |
377 | } |
378 | ||
2a0b4682 | 379 | refcount_set(&dd->count, 1); |
1da177e4 | 380 | list_add(&dd->list, &t->devices); |
afc567a4 | 381 | goto out; |
1da177e4 | 382 | |
86f1152b | 383 | } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { |
f165921d | 384 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 | 385 | if (r) |
f6007dce | 386 | goto unlock_ret_r; |
1da177e4 | 387 | } |
afc567a4 MS |
388 | refcount_inc(&dd->count); |
389 | out: | |
f6007dce | 390 | up_write(&t->devices_lock); |
86f1152b | 391 | *result = dd->dm_dev; |
1da177e4 | 392 | return 0; |
f6007dce MP |
393 | |
394 | unlock_ret_r: | |
395 | up_write(&t->devices_lock); | |
396 | return r; | |
1da177e4 | 397 | } |
08649012 | 398 | EXPORT_SYMBOL(dm_get_device); |
1da177e4 | 399 | |
11f0431b MS |
400 | static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
401 | sector_t start, sector_t len, void *data) | |
1da177e4 | 402 | { |
754c5fc7 MS |
403 | struct queue_limits *limits = data; |
404 | struct block_device *bdev = dev->bdev; | |
165125e1 | 405 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
406 | |
407 | if (unlikely(!q)) { | |
385411ff CH |
408 | DMWARN("%s: Cannot set limits for nonexistent device %pg", |
409 | dm_device_name(ti->table->md), bdev); | |
754c5fc7 | 410 | return 0; |
0c2322e4 | 411 | } |
3cb40214 | 412 | |
9efa82ef CH |
413 | if (blk_stack_limits(limits, &q->limits, |
414 | get_start_sect(bdev) + start) < 0) | |
385411ff | 415 | DMWARN("%s: adding target device %pg caused an alignment inconsistency: " |
a963a956 MS |
416 | "physical_block_size=%u, logical_block_size=%u, " |
417 | "alignment_offset=%u, start=%llu", | |
385411ff | 418 | dm_device_name(ti->table->md), bdev, |
a963a956 MS |
419 | q->limits.physical_block_size, |
420 | q->limits.logical_block_size, | |
421 | q->limits.alignment_offset, | |
b27d7f16 | 422 | (unsigned long long) start << SECTOR_SHIFT); |
754c5fc7 | 423 | return 0; |
3cb40214 | 424 | } |
969429b5 | 425 | |
1da177e4 | 426 | /* |
08649012 | 427 | * Decrement a device's use count and remove it if necessary. |
1da177e4 | 428 | */ |
82b1519b | 429 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 430 | { |
86f1152b | 431 | int found = 0; |
f6007dce MP |
432 | struct dm_table *t = ti->table; |
433 | struct list_head *devices = &t->devices; | |
86f1152b | 434 | struct dm_dev_internal *dd; |
82b1519b | 435 | |
f6007dce MP |
436 | down_write(&t->devices_lock); |
437 | ||
86f1152b BM |
438 | list_for_each_entry(dd, devices, list) { |
439 | if (dd->dm_dev == d) { | |
440 | found = 1; | |
441 | break; | |
442 | } | |
443 | } | |
444 | if (!found) { | |
43e6c111 | 445 | DMERR("%s: device %s not in table devices list", |
f6007dce MP |
446 | dm_device_name(t->md), d->name); |
447 | goto unlock_ret; | |
86f1152b | 448 | } |
2a0b4682 | 449 | if (refcount_dec_and_test(&dd->count)) { |
f6007dce | 450 | dm_put_table_device(t->md, d); |
1da177e4 LT |
451 | list_del(&dd->list); |
452 | kfree(dd); | |
453 | } | |
f6007dce MP |
454 | |
455 | unlock_ret: | |
456 | up_write(&t->devices_lock); | |
1da177e4 | 457 | } |
08649012 | 458 | EXPORT_SYMBOL(dm_put_device); |
1da177e4 LT |
459 | |
460 | /* | |
461 | * Checks to see if the target joins onto the end of the table. | |
462 | */ | |
564b5c54 | 463 | static int adjoin(struct dm_table *t, struct dm_target *ti) |
1da177e4 LT |
464 | { |
465 | struct dm_target *prev; | |
466 | ||
564b5c54 | 467 | if (!t->num_targets) |
1da177e4 LT |
468 | return !ti->begin; |
469 | ||
564b5c54 | 470 | prev = &t->targets[t->num_targets - 1]; |
1da177e4 LT |
471 | return (ti->begin == (prev->begin + prev->len)); |
472 | } | |
473 | ||
474 | /* | |
475 | * Used to dynamically allocate the arg array. | |
f36afb39 MP |
476 | * |
477 | * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must | |
478 | * process messages even if some device is suspended. These messages have a | |
479 | * small fixed number of arguments. | |
480 | * | |
481 | * On the other hand, dm-switch needs to process bulk data using messages and | |
482 | * excessive use of GFP_NOIO could cause trouble. | |
1da177e4 | 483 | */ |
86a3238c | 484 | static char **realloc_argv(unsigned int *size, char **old_argv) |
1da177e4 LT |
485 | { |
486 | char **argv; | |
86a3238c | 487 | unsigned int new_size; |
f36afb39 | 488 | gfp_t gfp; |
1da177e4 | 489 | |
610b15c5 KC |
490 | if (*size) { |
491 | new_size = *size * 2; | |
f36afb39 MP |
492 | gfp = GFP_KERNEL; |
493 | } else { | |
494 | new_size = 8; | |
495 | gfp = GFP_NOIO; | |
496 | } | |
6da2ec56 | 497 | argv = kmalloc_array(new_size, sizeof(*argv), gfp); |
a0651926 | 498 | if (argv && old_argv) { |
610b15c5 KC |
499 | memcpy(argv, old_argv, *size * sizeof(*argv)); |
500 | *size = new_size; | |
1da177e4 LT |
501 | } |
502 | ||
503 | kfree(old_argv); | |
504 | return argv; | |
505 | } | |
506 | ||
507 | /* | |
508 | * Destructively splits up the argument list to pass to ctr. | |
509 | */ | |
510 | int dm_split_args(int *argc, char ***argvp, char *input) | |
511 | { | |
512 | char *start, *end = input, *out, **argv = NULL; | |
86a3238c | 513 | unsigned int array_size = 0; |
1da177e4 LT |
514 | |
515 | *argc = 0; | |
814d6862 DT |
516 | |
517 | if (!input) { | |
518 | *argvp = NULL; | |
519 | return 0; | |
520 | } | |
521 | ||
1da177e4 LT |
522 | argv = realloc_argv(&array_size, argv); |
523 | if (!argv) | |
524 | return -ENOMEM; | |
525 | ||
526 | while (1) { | |
1da177e4 | 527 | /* Skip whitespace */ |
e7d2860b | 528 | start = skip_spaces(end); |
1da177e4 LT |
529 | |
530 | if (!*start) | |
531 | break; /* success, we hit the end */ | |
532 | ||
533 | /* 'out' is used to remove any back-quotes */ | |
534 | end = out = start; | |
535 | while (*end) { | |
536 | /* Everything apart from '\0' can be quoted */ | |
537 | if (*end == '\\' && *(end + 1)) { | |
538 | *out++ = *(end + 1); | |
539 | end += 2; | |
540 | continue; | |
541 | } | |
542 | ||
543 | if (isspace(*end)) | |
544 | break; /* end of token */ | |
545 | ||
546 | *out++ = *end++; | |
547 | } | |
548 | ||
549 | /* have we already filled the array ? */ | |
550 | if ((*argc + 1) > array_size) { | |
551 | argv = realloc_argv(&array_size, argv); | |
552 | if (!argv) | |
553 | return -ENOMEM; | |
554 | } | |
555 | ||
556 | /* we know this is whitespace */ | |
557 | if (*end) | |
558 | end++; | |
559 | ||
560 | /* terminate the string and put it in the array */ | |
561 | *out = '\0'; | |
562 | argv[*argc] = start; | |
563 | (*argc)++; | |
564 | } | |
565 | ||
566 | *argvp = argv; | |
567 | return 0; | |
568 | } | |
569 | ||
be6d4305 MS |
570 | /* |
571 | * Impose necessary and sufficient conditions on a devices's table such | |
572 | * that any incoming bio which respects its logical_block_size can be | |
573 | * processed successfully. If it falls across the boundary between | |
574 | * two or more targets, the size of each piece it gets split into must | |
575 | * be compatible with the logical_block_size of the target processing it. | |
576 | */ | |
564b5c54 MS |
577 | static int validate_hardware_logical_block_alignment(struct dm_table *t, |
578 | struct queue_limits *limits) | |
be6d4305 MS |
579 | { |
580 | /* | |
581 | * This function uses arithmetic modulo the logical_block_size | |
582 | * (in units of 512-byte sectors). | |
583 | */ | |
584 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 585 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
586 | |
587 | /* | |
588 | * Offset of the start of the next table entry, mod logical_block_size. | |
589 | */ | |
590 | unsigned short next_target_start = 0; | |
591 | ||
592 | /* | |
593 | * Given an aligned bio that extends beyond the end of a | |
594 | * target, how many sectors must the next target handle? | |
595 | */ | |
596 | unsigned short remaining = 0; | |
597 | ||
3f649ab7 | 598 | struct dm_target *ti; |
754c5fc7 | 599 | struct queue_limits ti_limits; |
564b5c54 | 600 | unsigned int i; |
be6d4305 MS |
601 | |
602 | /* | |
603 | * Check each entry in the table in turn. | |
604 | */ | |
564b5c54 MS |
605 | for (i = 0; i < t->num_targets; i++) { |
606 | ti = dm_table_get_target(t, i); | |
be6d4305 | 607 | |
b1bd055d | 608 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
609 | |
610 | /* combine all target devices' limits */ | |
611 | if (ti->type->iterate_devices) | |
612 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
613 | &ti_limits); | |
614 | ||
be6d4305 MS |
615 | /* |
616 | * If the remaining sectors fall entirely within this | |
617 | * table entry are they compatible with its logical_block_size? | |
618 | */ | |
619 | if (remaining < ti->len && | |
754c5fc7 | 620 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
621 | SECTOR_SHIFT) - 1)) |
622 | break; /* Error */ | |
623 | ||
624 | next_target_start = | |
625 | (unsigned short) ((next_target_start + ti->len) & | |
626 | (device_logical_block_size_sects - 1)); | |
627 | remaining = next_target_start ? | |
628 | device_logical_block_size_sects - next_target_start : 0; | |
629 | } | |
630 | ||
631 | if (remaining) { | |
43e6c111 MP |
632 | DMERR("%s: table line %u (start sect %llu len %llu) " |
633 | "not aligned to h/w logical block size %u", | |
634 | dm_device_name(t->md), i, | |
635 | (unsigned long long) ti->begin, | |
636 | (unsigned long long) ti->len, | |
637 | limits->logical_block_size); | |
be6d4305 MS |
638 | return -EINVAL; |
639 | } | |
640 | ||
641 | return 0; | |
642 | } | |
643 | ||
1da177e4 LT |
644 | int dm_table_add_target(struct dm_table *t, const char *type, |
645 | sector_t start, sector_t len, char *params) | |
646 | { | |
647 | int r = -EINVAL, argc; | |
648 | char **argv; | |
899ab445 | 649 | struct dm_target *ti; |
1da177e4 | 650 | |
3791e2fc AK |
651 | if (t->singleton) { |
652 | DMERR("%s: target type %s must appear alone in table", | |
653 | dm_device_name(t->md), t->targets->type->name); | |
654 | return -EINVAL; | |
655 | } | |
656 | ||
57a2f238 | 657 | BUG_ON(t->num_targets >= t->num_allocated); |
1da177e4 | 658 | |
899ab445 MS |
659 | ti = t->targets + t->num_targets; |
660 | memset(ti, 0, sizeof(*ti)); | |
1da177e4 LT |
661 | |
662 | if (!len) { | |
72d94861 | 663 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
664 | return -EINVAL; |
665 | } | |
666 | ||
899ab445 MS |
667 | ti->type = dm_get_target_type(type); |
668 | if (!ti->type) { | |
dafa724b | 669 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), type); |
1da177e4 LT |
670 | return -EINVAL; |
671 | } | |
672 | ||
899ab445 | 673 | if (dm_target_needs_singleton(ti->type)) { |
3791e2fc | 674 | if (t->num_targets) { |
899ab445 | 675 | ti->error = "singleton target type must appear alone in table"; |
dafa724b | 676 | goto bad; |
3791e2fc | 677 | } |
e83068a5 | 678 | t->singleton = true; |
3791e2fc AK |
679 | } |
680 | ||
05bdb996 CH |
681 | if (dm_target_always_writeable(ti->type) && |
682 | !(t->mode & BLK_OPEN_WRITE)) { | |
899ab445 | 683 | ti->error = "target type may not be included in a read-only table"; |
dafa724b | 684 | goto bad; |
cc6cbe14 AK |
685 | } |
686 | ||
36a0456f | 687 | if (t->immutable_target_type) { |
899ab445 MS |
688 | if (t->immutable_target_type != ti->type) { |
689 | ti->error = "immutable target type cannot be mixed with other target types"; | |
dafa724b | 690 | goto bad; |
36a0456f | 691 | } |
899ab445 | 692 | } else if (dm_target_is_immutable(ti->type)) { |
36a0456f | 693 | if (t->num_targets) { |
899ab445 | 694 | ti->error = "immutable target type cannot be mixed with other target types"; |
dafa724b | 695 | goto bad; |
36a0456f | 696 | } |
899ab445 | 697 | t->immutable_target_type = ti->type; |
36a0456f AK |
698 | } |
699 | ||
899ab445 | 700 | if (dm_target_has_integrity(ti->type)) |
9b4b5a79 MB |
701 | t->integrity_added = 1; |
702 | ||
899ab445 MS |
703 | ti->table = t; |
704 | ti->begin = start; | |
705 | ti->len = len; | |
706 | ti->error = "Unknown error"; | |
1da177e4 LT |
707 | |
708 | /* | |
709 | * Does this target adjoin the previous one ? | |
710 | */ | |
899ab445 MS |
711 | if (!adjoin(t, ti)) { |
712 | ti->error = "Gap in table"; | |
1da177e4 LT |
713 | goto bad; |
714 | } | |
715 | ||
716 | r = dm_split_args(&argc, &argv, params); | |
717 | if (r) { | |
899ab445 | 718 | ti->error = "couldn't split parameters"; |
1da177e4 LT |
719 | goto bad; |
720 | } | |
721 | ||
899ab445 | 722 | r = ti->type->ctr(ti, argc, argv); |
1da177e4 LT |
723 | kfree(argv); |
724 | if (r) | |
725 | goto bad; | |
726 | ||
899ab445 | 727 | t->highs[t->num_targets++] = ti->begin + ti->len - 1; |
1da177e4 | 728 | |
899ab445 | 729 | if (!ti->num_discard_bios && ti->discards_supported) |
55a62eef | 730 | DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.", |
936688d7 | 731 | dm_device_name(t->md), type); |
5ae89a87 | 732 | |
899ab445 | 733 | if (ti->limit_swap_bios && !static_key_enabled(&swap_bios_enabled.key)) |
442761fd MS |
734 | static_branch_enable(&swap_bios_enabled); |
735 | ||
1da177e4 LT |
736 | return 0; |
737 | ||
738 | bad: | |
899ab445 MS |
739 | DMERR("%s: %s: %s (%pe)", dm_device_name(t->md), type, ti->error, ERR_PTR(r)); |
740 | dm_put_target_type(ti->type); | |
1da177e4 LT |
741 | return r; |
742 | } | |
743 | ||
498f0103 MS |
744 | /* |
745 | * Target argument parsing helpers. | |
746 | */ | |
86a3238c HM |
747 | static int validate_next_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
748 | unsigned int *value, char **error, unsigned int grouped) | |
498f0103 MS |
749 | { |
750 | const char *arg_str = dm_shift_arg(arg_set); | |
31998ef1 | 751 | char dummy; |
498f0103 MS |
752 | |
753 | if (!arg_str || | |
31998ef1 | 754 | (sscanf(arg_str, "%u%c", value, &dummy) != 1) || |
498f0103 MS |
755 | (*value < arg->min) || |
756 | (*value > arg->max) || | |
757 | (grouped && arg_set->argc < *value)) { | |
758 | *error = arg->error; | |
759 | return -EINVAL; | |
760 | } | |
761 | ||
762 | return 0; | |
763 | } | |
764 | ||
5916a22b | 765 | int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
86a3238c | 766 | unsigned int *value, char **error) |
498f0103 MS |
767 | { |
768 | return validate_next_arg(arg, arg_set, value, error, 0); | |
769 | } | |
770 | EXPORT_SYMBOL(dm_read_arg); | |
771 | ||
5916a22b | 772 | int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
86a3238c | 773 | unsigned int *value, char **error) |
498f0103 MS |
774 | { |
775 | return validate_next_arg(arg, arg_set, value, error, 1); | |
776 | } | |
777 | EXPORT_SYMBOL(dm_read_arg_group); | |
778 | ||
779 | const char *dm_shift_arg(struct dm_arg_set *as) | |
780 | { | |
781 | char *r; | |
782 | ||
783 | if (as->argc) { | |
784 | as->argc--; | |
785 | r = *as->argv; | |
786 | as->argv++; | |
787 | return r; | |
788 | } | |
789 | ||
790 | return NULL; | |
791 | } | |
792 | EXPORT_SYMBOL(dm_shift_arg); | |
793 | ||
86a3238c | 794 | void dm_consume_args(struct dm_arg_set *as, unsigned int num_args) |
498f0103 MS |
795 | { |
796 | BUG_ON(as->argc < num_args); | |
797 | as->argc -= num_args; | |
798 | as->argv += num_args; | |
799 | } | |
800 | EXPORT_SYMBOL(dm_consume_args); | |
801 | ||
7e0d574f | 802 | static bool __table_type_bio_based(enum dm_queue_mode table_type) |
545ed20e TK |
803 | { |
804 | return (table_type == DM_TYPE_BIO_BASED || | |
9c37de29 | 805 | table_type == DM_TYPE_DAX_BIO_BASED); |
545ed20e TK |
806 | } |
807 | ||
7e0d574f | 808 | static bool __table_type_request_based(enum dm_queue_mode table_type) |
15b94a69 | 809 | { |
953923c0 | 810 | return table_type == DM_TYPE_REQUEST_BASED; |
15b94a69 JN |
811 | } |
812 | ||
7e0d574f | 813 | void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type) |
e83068a5 MS |
814 | { |
815 | t->type = type; | |
816 | } | |
817 | EXPORT_SYMBOL_GPL(dm_table_set_type); | |
818 | ||
7bf7eac8 | 819 | /* validate the dax capability of the target device span */ |
7b0800d0 | 820 | static int device_not_dax_capable(struct dm_target *ti, struct dm_dev *dev, |
9c50a98f | 821 | sector_t start, sector_t len, void *data) |
545ed20e | 822 | { |
7b0800d0 CH |
823 | if (dev->dax_dev) |
824 | return false; | |
7bf7eac8 | 825 | |
7b0800d0 CH |
826 | DMDEBUG("%pg: error: dax unsupported by block device", dev->bdev); |
827 | return true; | |
545ed20e TK |
828 | } |
829 | ||
2e9ee095 | 830 | /* Check devices support synchronous DAX */ |
5b0fab50 JX |
831 | static int device_not_dax_synchronous_capable(struct dm_target *ti, struct dm_dev *dev, |
832 | sector_t start, sector_t len, void *data) | |
2e9ee095 | 833 | { |
5b0fab50 | 834 | return !dev->dax_dev || !dax_synchronous(dev->dax_dev); |
2e9ee095 PG |
835 | } |
836 | ||
7b0800d0 | 837 | static bool dm_table_supports_dax(struct dm_table *t, |
564b5c54 | 838 | iterate_devices_callout_fn iterate_fn) |
545ed20e | 839 | { |
545ed20e | 840 | /* Ensure that all targets support DAX. */ |
564b5c54 MS |
841 | for (unsigned int i = 0; i < t->num_targets; i++) { |
842 | struct dm_target *ti = dm_table_get_target(t, i); | |
545ed20e TK |
843 | |
844 | if (!ti->type->direct_access) | |
845 | return false; | |
846 | ||
847 | if (!ti->type->iterate_devices || | |
7b0800d0 | 848 | ti->type->iterate_devices(ti, iterate_fn, NULL)) |
545ed20e TK |
849 | return false; |
850 | } | |
851 | ||
852 | return true; | |
853 | } | |
854 | ||
6ba01df7 MS |
855 | static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev, |
856 | sector_t start, sector_t len, void *data) | |
eaa160ed | 857 | { |
6ba01df7 MS |
858 | struct block_device *bdev = dev->bdev; |
859 | struct request_queue *q = bdev_get_queue(bdev); | |
eaa160ed | 860 | |
6ba01df7 | 861 | /* request-based cannot stack on partitions! */ |
fa01b1e9 | 862 | if (bdev_is_partition(bdev)) |
6ba01df7 | 863 | return false; |
eaa160ed | 864 | |
344e9ffc | 865 | return queue_is_mq(q); |
eaa160ed MS |
866 | } |
867 | ||
e83068a5 | 868 | static int dm_table_determine_type(struct dm_table *t) |
e6ee8c0b | 869 | { |
86a3238c | 870 | unsigned int bio_based = 0, request_based = 0, hybrid = 0; |
564b5c54 | 871 | struct dm_target *ti; |
e83068a5 | 872 | struct list_head *devices = dm_table_get_devices(t); |
7e0d574f | 873 | enum dm_queue_mode live_md_type = dm_get_md_type(t->md); |
e6ee8c0b | 874 | |
e83068a5 MS |
875 | if (t->type != DM_TYPE_NONE) { |
876 | /* target already set the table's type */ | |
c934edad MS |
877 | if (t->type == DM_TYPE_BIO_BASED) { |
878 | /* possibly upgrade to a variant of bio-based */ | |
879 | goto verify_bio_based; | |
22c11858 | 880 | } |
545ed20e | 881 | BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED); |
e83068a5 MS |
882 | goto verify_rq_based; |
883 | } | |
884 | ||
564b5c54 MS |
885 | for (unsigned int i = 0; i < t->num_targets; i++) { |
886 | ti = dm_table_get_target(t, i); | |
887 | if (dm_target_hybrid(ti)) | |
169e2cc2 | 888 | hybrid = 1; |
564b5c54 | 889 | else if (dm_target_request_based(ti)) |
e6ee8c0b KU |
890 | request_based = 1; |
891 | else | |
892 | bio_based = 1; | |
893 | ||
894 | if (bio_based && request_based) { | |
2e84fecf | 895 | DMERR("Inconsistent table: different target types can't be mixed up"); |
e6ee8c0b KU |
896 | return -EINVAL; |
897 | } | |
898 | } | |
899 | ||
169e2cc2 MS |
900 | if (hybrid && !bio_based && !request_based) { |
901 | /* | |
902 | * The targets can work either way. | |
903 | * Determine the type from the live device. | |
904 | * Default to bio-based if device is new. | |
905 | */ | |
15b94a69 | 906 | if (__table_type_request_based(live_md_type)) |
169e2cc2 MS |
907 | request_based = 1; |
908 | else | |
909 | bio_based = 1; | |
910 | } | |
911 | ||
e6ee8c0b | 912 | if (bio_based) { |
c934edad | 913 | verify_bio_based: |
e6ee8c0b KU |
914 | /* We must use this table as bio-based */ |
915 | t->type = DM_TYPE_BIO_BASED; | |
7b0800d0 | 916 | if (dm_table_supports_dax(t, device_not_dax_capable) || |
22c11858 | 917 | (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) { |
545ed20e | 918 | t->type = DM_TYPE_DAX_BIO_BASED; |
22c11858 | 919 | } |
e6ee8c0b KU |
920 | return 0; |
921 | } | |
922 | ||
923 | BUG_ON(!request_based); /* No targets in this table */ | |
924 | ||
e83068a5 MS |
925 | t->type = DM_TYPE_REQUEST_BASED; |
926 | ||
927 | verify_rq_based: | |
65803c20 MS |
928 | /* |
929 | * Request-based dm supports only tables that have a single target now. | |
930 | * To support multiple targets, request splitting support is needed, | |
931 | * and that needs lots of changes in the block-layer. | |
932 | * (e.g. request completion process for partial completion.) | |
933 | */ | |
934 | if (t->num_targets > 1) { | |
9c37de29 | 935 | DMERR("request-based DM doesn't support multiple targets"); |
65803c20 MS |
936 | return -EINVAL; |
937 | } | |
938 | ||
6936c12c MS |
939 | if (list_empty(devices)) { |
940 | int srcu_idx; | |
941 | struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx); | |
942 | ||
6a23e05c JA |
943 | /* inherit live table's type */ |
944 | if (live_table) | |
6936c12c | 945 | t->type = live_table->type; |
6936c12c MS |
946 | dm_put_live_table(t->md, srcu_idx); |
947 | return 0; | |
948 | } | |
949 | ||
564b5c54 MS |
950 | ti = dm_table_get_immutable_target(t); |
951 | if (!ti) { | |
22c11858 MS |
952 | DMERR("table load rejected: immutable target is required"); |
953 | return -EINVAL; | |
564b5c54 | 954 | } else if (ti->max_io_len) { |
22c11858 MS |
955 | DMERR("table load rejected: immutable target that splits IO is not supported"); |
956 | return -EINVAL; | |
957 | } | |
958 | ||
e6ee8c0b | 959 | /* Non-request-stackable devices can't be used for request-based dm */ |
564b5c54 MS |
960 | if (!ti->type->iterate_devices || |
961 | !ti->type->iterate_devices(ti, device_is_rq_stackable, NULL)) { | |
eaa160ed MS |
962 | DMERR("table load rejected: including non-request-stackable devices"); |
963 | return -EINVAL; | |
e5863d9a | 964 | } |
301fc3f5 | 965 | |
e6ee8c0b KU |
966 | return 0; |
967 | } | |
968 | ||
7e0d574f | 969 | enum dm_queue_mode dm_table_get_type(struct dm_table *t) |
e6ee8c0b KU |
970 | { |
971 | return t->type; | |
972 | } | |
973 | ||
36a0456f AK |
974 | struct target_type *dm_table_get_immutable_target_type(struct dm_table *t) |
975 | { | |
976 | return t->immutable_target_type; | |
977 | } | |
978 | ||
16f12266 MS |
979 | struct dm_target *dm_table_get_immutable_target(struct dm_table *t) |
980 | { | |
981 | /* Immutable target is implicitly a singleton */ | |
982 | if (t->num_targets > 1 || | |
983 | !dm_target_is_immutable(t->targets[0].type)) | |
984 | return NULL; | |
985 | ||
986 | return t->targets; | |
987 | } | |
988 | ||
f083b09b MS |
989 | struct dm_target *dm_table_get_wildcard_target(struct dm_table *t) |
990 | { | |
564b5c54 MS |
991 | for (unsigned int i = 0; i < t->num_targets; i++) { |
992 | struct dm_target *ti = dm_table_get_target(t, i); | |
f083b09b | 993 | |
f083b09b MS |
994 | if (dm_target_is_wildcard(ti->type)) |
995 | return ti; | |
996 | } | |
997 | ||
998 | return NULL; | |
999 | } | |
1000 | ||
545ed20e TK |
1001 | bool dm_table_bio_based(struct dm_table *t) |
1002 | { | |
1003 | return __table_type_bio_based(dm_table_get_type(t)); | |
1004 | } | |
1005 | ||
e6ee8c0b KU |
1006 | bool dm_table_request_based(struct dm_table *t) |
1007 | { | |
15b94a69 | 1008 | return __table_type_request_based(dm_table_get_type(t)); |
e5863d9a MS |
1009 | } |
1010 | ||
9571f829 | 1011 | static bool dm_table_supports_poll(struct dm_table *t); |
cfc97abc | 1012 | |
17e149b8 | 1013 | static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md) |
e6ee8c0b | 1014 | { |
7e0d574f | 1015 | enum dm_queue_mode type = dm_table_get_type(t); |
e810cb78 CH |
1016 | unsigned int per_io_data_size = 0, front_pad, io_front_pad; |
1017 | unsigned int min_pool_size = 0, pool_size; | |
1018 | struct dm_md_mempools *pools; | |
e6ee8c0b | 1019 | |
78d8e58a | 1020 | if (unlikely(type == DM_TYPE_NONE)) { |
43e6c111 | 1021 | DMERR("no table type is set, can't allocate mempools"); |
e6ee8c0b KU |
1022 | return -EINVAL; |
1023 | } | |
1024 | ||
e810cb78 CH |
1025 | pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id); |
1026 | if (!pools) | |
1027 | return -ENOMEM; | |
1028 | ||
1029 | if (type == DM_TYPE_REQUEST_BASED) { | |
1030 | pool_size = dm_get_reserved_rq_based_ios(); | |
1031 | front_pad = offsetof(struct dm_rq_clone_bio_info, clone); | |
1032 | goto init_bs; | |
cfc97abc | 1033 | } |
78d8e58a | 1034 | |
e810cb78 | 1035 | for (unsigned int i = 0; i < t->num_targets; i++) { |
564b5c54 | 1036 | struct dm_target *ti = dm_table_get_target(t, i); |
e6ee8c0b | 1037 | |
e810cb78 CH |
1038 | per_io_data_size = max(per_io_data_size, ti->per_io_data_size); |
1039 | min_pool_size = max(min_pool_size, ti->num_flush_bios); | |
1040 | } | |
1041 | pool_size = max(dm_get_reserved_bio_based_ios(), min_pool_size); | |
1042 | front_pad = roundup(per_io_data_size, | |
1043 | __alignof__(struct dm_target_io)) + DM_TARGET_IO_BIO_OFFSET; | |
1044 | ||
1045 | io_front_pad = roundup(per_io_data_size, | |
1046 | __alignof__(struct dm_io)) + DM_IO_BIO_OFFSET; | |
1047 | if (bioset_init(&pools->io_bs, pool_size, io_front_pad, | |
1048 | dm_table_supports_poll(t) ? BIOSET_PERCPU_CACHE : 0)) | |
1049 | goto out_free_pools; | |
1050 | if (t->integrity_supported && | |
1051 | bioset_integrity_create(&pools->io_bs, pool_size)) | |
1052 | goto out_free_pools; | |
1053 | init_bs: | |
1054 | if (bioset_init(&pools->bs, pool_size, front_pad, 0)) | |
1055 | goto out_free_pools; | |
1056 | if (t->integrity_supported && | |
1057 | bioset_integrity_create(&pools->bs, pool_size)) | |
1058 | goto out_free_pools; | |
1059 | ||
1060 | t->mempools = pools; | |
e6ee8c0b | 1061 | return 0; |
e810cb78 CH |
1062 | |
1063 | out_free_pools: | |
1064 | dm_free_md_mempools(pools); | |
1065 | return -ENOMEM; | |
e6ee8c0b KU |
1066 | } |
1067 | ||
1da177e4 LT |
1068 | static int setup_indexes(struct dm_table *t) |
1069 | { | |
1070 | int i; | |
1071 | unsigned int total = 0; | |
1072 | sector_t *indexes; | |
1073 | ||
1074 | /* allocate the space for *all* the indexes */ | |
1075 | for (i = t->depth - 2; i >= 0; i--) { | |
1076 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
1077 | total += t->counts[i]; | |
1078 | } | |
1079 | ||
7a35693a | 1080 | indexes = kvcalloc(total, NODE_SIZE, GFP_KERNEL); |
1da177e4 LT |
1081 | if (!indexes) |
1082 | return -ENOMEM; | |
1083 | ||
1084 | /* set up internal nodes, bottom-up */ | |
82d601dc | 1085 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
1086 | t->index[i] = indexes; |
1087 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
1088 | setup_btree_index(i, t); | |
1089 | } | |
1090 | ||
1091 | return 0; | |
1092 | } | |
1093 | ||
1094 | /* | |
1095 | * Builds the btree to index the map. | |
1096 | */ | |
26803b9f | 1097 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
1098 | { |
1099 | int r = 0; | |
1100 | unsigned int leaf_nodes; | |
1101 | ||
1da177e4 LT |
1102 | /* how many indexes will the btree have ? */ |
1103 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
1104 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
1105 | ||
1106 | /* leaf layer has already been set up */ | |
1107 | t->counts[t->depth - 1] = leaf_nodes; | |
1108 | t->index[t->depth - 1] = t->highs; | |
1109 | ||
1110 | if (t->depth >= 2) | |
1111 | r = setup_indexes(t); | |
1112 | ||
1113 | return r; | |
1114 | } | |
1115 | ||
25520d55 MP |
1116 | static bool integrity_profile_exists(struct gendisk *disk) |
1117 | { | |
1118 | return !!blk_get_integrity(disk); | |
1119 | } | |
1120 | ||
a63a5cf8 MS |
1121 | /* |
1122 | * Get a disk whose integrity profile reflects the table's profile. | |
a63a5cf8 MS |
1123 | * Returns NULL if integrity support was inconsistent or unavailable. |
1124 | */ | |
43e6c111 | 1125 | static struct gendisk *dm_table_get_integrity_disk(struct dm_table *t) |
a63a5cf8 MS |
1126 | { |
1127 | struct list_head *devices = dm_table_get_devices(t); | |
1128 | struct dm_dev_internal *dd = NULL; | |
1129 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
e2460f2a | 1130 | |
564b5c54 | 1131 | for (unsigned int i = 0; i < t->num_targets; i++) { |
e2460f2a | 1132 | struct dm_target *ti = dm_table_get_target(t, i); |
564b5c54 | 1133 | |
e2460f2a MP |
1134 | if (!dm_target_passes_integrity(ti->type)) |
1135 | goto no_integrity; | |
1136 | } | |
a63a5cf8 MS |
1137 | |
1138 | list_for_each_entry(dd, devices, list) { | |
86f1152b | 1139 | template_disk = dd->dm_dev->bdev->bd_disk; |
25520d55 | 1140 | if (!integrity_profile_exists(template_disk)) |
a63a5cf8 | 1141 | goto no_integrity; |
a63a5cf8 MS |
1142 | else if (prev_disk && |
1143 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
1144 | goto no_integrity; | |
1145 | prev_disk = template_disk; | |
1146 | } | |
1147 | ||
1148 | return template_disk; | |
1149 | ||
1150 | no_integrity: | |
1151 | if (prev_disk) | |
1152 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
1153 | dm_device_name(t->md), | |
1154 | prev_disk->disk_name, | |
1155 | template_disk->disk_name); | |
1156 | return NULL; | |
1157 | } | |
1158 | ||
26803b9f | 1159 | /* |
25520d55 MP |
1160 | * Register the mapped device for blk_integrity support if the |
1161 | * underlying devices have an integrity profile. But all devices may | |
1162 | * not have matching profiles (checking all devices isn't reliable | |
a63a5cf8 | 1163 | * during table load because this table may use other DM device(s) which |
25520d55 MP |
1164 | * must be resumed before they will have an initialized integity |
1165 | * profile). Consequently, stacked DM devices force a 2 stage integrity | |
1166 | * profile validation: First pass during table load, final pass during | |
1167 | * resume. | |
26803b9f | 1168 | */ |
25520d55 | 1169 | static int dm_table_register_integrity(struct dm_table *t) |
26803b9f | 1170 | { |
25520d55 | 1171 | struct mapped_device *md = t->md; |
a63a5cf8 | 1172 | struct gendisk *template_disk = NULL; |
26803b9f | 1173 | |
9b4b5a79 MB |
1174 | /* If target handles integrity itself do not register it here. */ |
1175 | if (t->integrity_added) | |
1176 | return 0; | |
1177 | ||
25520d55 | 1178 | template_disk = dm_table_get_integrity_disk(t); |
a63a5cf8 MS |
1179 | if (!template_disk) |
1180 | return 0; | |
26803b9f | 1181 | |
25520d55 | 1182 | if (!integrity_profile_exists(dm_disk(md))) { |
e83068a5 | 1183 | t->integrity_supported = true; |
25520d55 MP |
1184 | /* |
1185 | * Register integrity profile during table load; we can do | |
1186 | * this because the final profile must match during resume. | |
1187 | */ | |
1188 | blk_integrity_register(dm_disk(md), | |
1189 | blk_get_integrity(template_disk)); | |
1190 | return 0; | |
a63a5cf8 MS |
1191 | } |
1192 | ||
1193 | /* | |
25520d55 | 1194 | * If DM device already has an initialized integrity |
a63a5cf8 MS |
1195 | * profile the new profile should not conflict. |
1196 | */ | |
25520d55 | 1197 | if (blk_integrity_compare(dm_disk(md), template_disk) < 0) { |
2e84fecf | 1198 | DMERR("%s: conflict with existing integrity profile: %s profile mismatch", |
43e6c111 MP |
1199 | dm_device_name(t->md), |
1200 | template_disk->disk_name); | |
a63a5cf8 MS |
1201 | return 1; |
1202 | } | |
1203 | ||
25520d55 | 1204 | /* Preserve existing integrity profile */ |
e83068a5 | 1205 | t->integrity_supported = true; |
26803b9f WD |
1206 | return 0; |
1207 | } | |
1208 | ||
aa6ce87a ST |
1209 | #ifdef CONFIG_BLK_INLINE_ENCRYPTION |
1210 | ||
cb77cb5a EB |
1211 | struct dm_crypto_profile { |
1212 | struct blk_crypto_profile profile; | |
aa6ce87a ST |
1213 | struct mapped_device *md; |
1214 | }; | |
1215 | ||
9355a9eb ST |
1216 | static int dm_keyslot_evict_callback(struct dm_target *ti, struct dm_dev *dev, |
1217 | sector_t start, sector_t len, void *data) | |
1218 | { | |
70493a63 | 1219 | const struct blk_crypto_key *key = data; |
9355a9eb | 1220 | |
70493a63 | 1221 | blk_crypto_evict_key(dev->bdev, key); |
9355a9eb ST |
1222 | return 0; |
1223 | } | |
1224 | ||
1225 | /* | |
1226 | * When an inline encryption key is evicted from a device-mapper device, evict | |
1227 | * it from all the underlying devices. | |
1228 | */ | |
cb77cb5a | 1229 | static int dm_keyslot_evict(struct blk_crypto_profile *profile, |
9355a9eb ST |
1230 | const struct blk_crypto_key *key, unsigned int slot) |
1231 | { | |
cb77cb5a EB |
1232 | struct mapped_device *md = |
1233 | container_of(profile, struct dm_crypto_profile, profile)->md; | |
9355a9eb ST |
1234 | struct dm_table *t; |
1235 | int srcu_idx; | |
9355a9eb ST |
1236 | |
1237 | t = dm_get_live_table(md, &srcu_idx); | |
1238 | if (!t) | |
1239 | return 0; | |
564b5c54 MS |
1240 | |
1241 | for (unsigned int i = 0; i < t->num_targets; i++) { | |
1242 | struct dm_target *ti = dm_table_get_target(t, i); | |
1243 | ||
9355a9eb ST |
1244 | if (!ti->type->iterate_devices) |
1245 | continue; | |
70493a63 EB |
1246 | ti->type->iterate_devices(ti, dm_keyslot_evict_callback, |
1247 | (void *)key); | |
9355a9eb | 1248 | } |
564b5c54 | 1249 | |
9355a9eb | 1250 | dm_put_live_table(md, srcu_idx); |
70493a63 | 1251 | return 0; |
9355a9eb ST |
1252 | } |
1253 | ||
cb77cb5a EB |
1254 | static int |
1255 | device_intersect_crypto_capabilities(struct dm_target *ti, struct dm_dev *dev, | |
1256 | sector_t start, sector_t len, void *data) | |
aa6ce87a | 1257 | { |
cb77cb5a EB |
1258 | struct blk_crypto_profile *parent = data; |
1259 | struct blk_crypto_profile *child = | |
1260 | bdev_get_queue(dev->bdev)->crypto_profile; | |
aa6ce87a | 1261 | |
cb77cb5a | 1262 | blk_crypto_intersect_capabilities(parent, child); |
aa6ce87a ST |
1263 | return 0; |
1264 | } | |
1265 | ||
cb77cb5a | 1266 | void dm_destroy_crypto_profile(struct blk_crypto_profile *profile) |
aa6ce87a | 1267 | { |
cb77cb5a EB |
1268 | struct dm_crypto_profile *dmcp = container_of(profile, |
1269 | struct dm_crypto_profile, | |
1270 | profile); | |
aa6ce87a | 1271 | |
cb77cb5a | 1272 | if (!profile) |
aa6ce87a ST |
1273 | return; |
1274 | ||
cb77cb5a EB |
1275 | blk_crypto_profile_destroy(profile); |
1276 | kfree(dmcp); | |
aa6ce87a ST |
1277 | } |
1278 | ||
cb77cb5a | 1279 | static void dm_table_destroy_crypto_profile(struct dm_table *t) |
aa6ce87a | 1280 | { |
cb77cb5a EB |
1281 | dm_destroy_crypto_profile(t->crypto_profile); |
1282 | t->crypto_profile = NULL; | |
aa6ce87a ST |
1283 | } |
1284 | ||
1285 | /* | |
cb77cb5a EB |
1286 | * Constructs and initializes t->crypto_profile with a crypto profile that |
1287 | * represents the common set of crypto capabilities of the devices described by | |
1288 | * the dm_table. However, if the constructed crypto profile doesn't support all | |
1289 | * crypto capabilities that are supported by the current mapped_device, it | |
1290 | * returns an error instead, since we don't support removing crypto capabilities | |
1291 | * on table changes. Finally, if the constructed crypto profile is "empty" (has | |
1292 | * no crypto capabilities at all), it just sets t->crypto_profile to NULL. | |
aa6ce87a | 1293 | */ |
cb77cb5a | 1294 | static int dm_table_construct_crypto_profile(struct dm_table *t) |
aa6ce87a | 1295 | { |
cb77cb5a EB |
1296 | struct dm_crypto_profile *dmcp; |
1297 | struct blk_crypto_profile *profile; | |
aa6ce87a | 1298 | unsigned int i; |
cb77cb5a | 1299 | bool empty_profile = true; |
aa6ce87a | 1300 | |
cb77cb5a EB |
1301 | dmcp = kmalloc(sizeof(*dmcp), GFP_KERNEL); |
1302 | if (!dmcp) | |
aa6ce87a | 1303 | return -ENOMEM; |
cb77cb5a | 1304 | dmcp->md = t->md; |
aa6ce87a | 1305 | |
cb77cb5a EB |
1306 | profile = &dmcp->profile; |
1307 | blk_crypto_profile_init(profile, 0); | |
1308 | profile->ll_ops.keyslot_evict = dm_keyslot_evict; | |
1309 | profile->max_dun_bytes_supported = UINT_MAX; | |
1310 | memset(profile->modes_supported, 0xFF, | |
1311 | sizeof(profile->modes_supported)); | |
aa6ce87a | 1312 | |
2aec377a | 1313 | for (i = 0; i < t->num_targets; i++) { |
564b5c54 | 1314 | struct dm_target *ti = dm_table_get_target(t, i); |
aa6ce87a ST |
1315 | |
1316 | if (!dm_target_passes_crypto(ti->type)) { | |
cb77cb5a | 1317 | blk_crypto_intersect_capabilities(profile, NULL); |
aa6ce87a ST |
1318 | break; |
1319 | } | |
1320 | if (!ti->type->iterate_devices) | |
1321 | continue; | |
cb77cb5a EB |
1322 | ti->type->iterate_devices(ti, |
1323 | device_intersect_crypto_capabilities, | |
1324 | profile); | |
aa6ce87a ST |
1325 | } |
1326 | ||
cb77cb5a EB |
1327 | if (t->md->queue && |
1328 | !blk_crypto_has_capabilities(profile, | |
1329 | t->md->queue->crypto_profile)) { | |
43e6c111 | 1330 | DMERR("Inline encryption capabilities of new DM table were more restrictive than the old table's. This is not supported!"); |
cb77cb5a | 1331 | dm_destroy_crypto_profile(profile); |
aa6ce87a ST |
1332 | return -EINVAL; |
1333 | } | |
1334 | ||
1335 | /* | |
cb77cb5a EB |
1336 | * If the new profile doesn't actually support any crypto capabilities, |
1337 | * we may as well represent it with a NULL profile. | |
aa6ce87a | 1338 | */ |
cb77cb5a EB |
1339 | for (i = 0; i < ARRAY_SIZE(profile->modes_supported); i++) { |
1340 | if (profile->modes_supported[i]) { | |
1341 | empty_profile = false; | |
aa6ce87a ST |
1342 | break; |
1343 | } | |
1344 | } | |
1345 | ||
cb77cb5a EB |
1346 | if (empty_profile) { |
1347 | dm_destroy_crypto_profile(profile); | |
1348 | profile = NULL; | |
aa6ce87a ST |
1349 | } |
1350 | ||
1351 | /* | |
cb77cb5a EB |
1352 | * t->crypto_profile is only set temporarily while the table is being |
1353 | * set up, and it gets set to NULL after the profile has been | |
1354 | * transferred to the request_queue. | |
aa6ce87a | 1355 | */ |
cb77cb5a | 1356 | t->crypto_profile = profile; |
aa6ce87a ST |
1357 | |
1358 | return 0; | |
1359 | } | |
1360 | ||
cb77cb5a EB |
1361 | static void dm_update_crypto_profile(struct request_queue *q, |
1362 | struct dm_table *t) | |
aa6ce87a | 1363 | { |
cb77cb5a | 1364 | if (!t->crypto_profile) |
aa6ce87a ST |
1365 | return; |
1366 | ||
cb77cb5a EB |
1367 | /* Make the crypto profile less restrictive. */ |
1368 | if (!q->crypto_profile) { | |
1369 | blk_crypto_register(t->crypto_profile, q); | |
aa6ce87a | 1370 | } else { |
cb77cb5a EB |
1371 | blk_crypto_update_capabilities(q->crypto_profile, |
1372 | t->crypto_profile); | |
1373 | dm_destroy_crypto_profile(t->crypto_profile); | |
aa6ce87a | 1374 | } |
cb77cb5a | 1375 | t->crypto_profile = NULL; |
aa6ce87a ST |
1376 | } |
1377 | ||
1378 | #else /* CONFIG_BLK_INLINE_ENCRYPTION */ | |
1379 | ||
cb77cb5a | 1380 | static int dm_table_construct_crypto_profile(struct dm_table *t) |
aa6ce87a ST |
1381 | { |
1382 | return 0; | |
1383 | } | |
1384 | ||
cb77cb5a | 1385 | void dm_destroy_crypto_profile(struct blk_crypto_profile *profile) |
aa6ce87a ST |
1386 | { |
1387 | } | |
1388 | ||
cb77cb5a | 1389 | static void dm_table_destroy_crypto_profile(struct dm_table *t) |
aa6ce87a ST |
1390 | { |
1391 | } | |
1392 | ||
cb77cb5a EB |
1393 | static void dm_update_crypto_profile(struct request_queue *q, |
1394 | struct dm_table *t) | |
aa6ce87a ST |
1395 | { |
1396 | } | |
1397 | ||
1398 | #endif /* !CONFIG_BLK_INLINE_ENCRYPTION */ | |
1399 | ||
26803b9f WD |
1400 | /* |
1401 | * Prepares the table for use by building the indices, | |
1402 | * setting the type, and allocating mempools. | |
1403 | */ | |
1404 | int dm_table_complete(struct dm_table *t) | |
1405 | { | |
1406 | int r; | |
1407 | ||
e83068a5 | 1408 | r = dm_table_determine_type(t); |
26803b9f | 1409 | if (r) { |
e83068a5 | 1410 | DMERR("unable to determine table type"); |
26803b9f WD |
1411 | return r; |
1412 | } | |
1413 | ||
1414 | r = dm_table_build_index(t); | |
1415 | if (r) { | |
1416 | DMERR("unable to build btrees"); | |
1417 | return r; | |
1418 | } | |
1419 | ||
25520d55 | 1420 | r = dm_table_register_integrity(t); |
26803b9f WD |
1421 | if (r) { |
1422 | DMERR("could not register integrity profile."); | |
1423 | return r; | |
1424 | } | |
1425 | ||
cb77cb5a | 1426 | r = dm_table_construct_crypto_profile(t); |
aa6ce87a | 1427 | if (r) { |
cb77cb5a | 1428 | DMERR("could not construct crypto profile."); |
aa6ce87a ST |
1429 | return r; |
1430 | } | |
1431 | ||
17e149b8 | 1432 | r = dm_table_alloc_md_mempools(t, t->md); |
26803b9f WD |
1433 | if (r) |
1434 | DMERR("unable to allocate mempools"); | |
1435 | ||
1436 | return r; | |
1437 | } | |
1438 | ||
48c9c27b | 1439 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1440 | void dm_table_event_callback(struct dm_table *t, |
1441 | void (*fn)(void *), void *context) | |
1442 | { | |
48c9c27b | 1443 | mutex_lock(&_event_lock); |
1da177e4 LT |
1444 | t->event_fn = fn; |
1445 | t->event_context = context; | |
48c9c27b | 1446 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1447 | } |
1448 | ||
1449 | void dm_table_event(struct dm_table *t) | |
1450 | { | |
48c9c27b | 1451 | mutex_lock(&_event_lock); |
1da177e4 LT |
1452 | if (t->event_fn) |
1453 | t->event_fn(t->event_context); | |
48c9c27b | 1454 | mutex_unlock(&_event_lock); |
1da177e4 | 1455 | } |
08649012 | 1456 | EXPORT_SYMBOL(dm_table_event); |
1da177e4 | 1457 | |
1cfd5d33 | 1458 | inline sector_t dm_table_get_size(struct dm_table *t) |
1da177e4 LT |
1459 | { |
1460 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1461 | } | |
08649012 | 1462 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1463 | |
1da177e4 LT |
1464 | /* |
1465 | * Search the btree for the correct target. | |
512875bd | 1466 | * |
123d87d5 | 1467 | * Caller should check returned pointer for NULL |
512875bd | 1468 | * to trap I/O beyond end of device. |
1da177e4 LT |
1469 | */ |
1470 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1471 | { | |
1472 | unsigned int l, n = 0, k = 0; | |
1473 | sector_t *node; | |
1474 | ||
1cfd5d33 | 1475 | if (unlikely(sector >= dm_table_get_size(t))) |
123d87d5 | 1476 | return NULL; |
1cfd5d33 | 1477 | |
1da177e4 LT |
1478 | for (l = 0; l < t->depth; l++) { |
1479 | n = get_child(n, k); | |
1480 | node = get_node(t, l, n); | |
1481 | ||
1482 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1483 | if (node[k] >= sector) | |
1484 | break; | |
1485 | } | |
1486 | ||
1487 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1488 | } | |
1489 | ||
b99fdcdc ML |
1490 | static int device_not_poll_capable(struct dm_target *ti, struct dm_dev *dev, |
1491 | sector_t start, sector_t len, void *data) | |
1492 | { | |
1493 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1494 | ||
1495 | return !test_bit(QUEUE_FLAG_POLL, &q->queue_flags); | |
1496 | } | |
1497 | ||
a4c8dd9c JX |
1498 | /* |
1499 | * type->iterate_devices() should be called when the sanity check needs to | |
1500 | * iterate and check all underlying data devices. iterate_devices() will | |
1501 | * iterate all underlying data devices until it encounters a non-zero return | |
1502 | * code, returned by whether the input iterate_devices_callout_fn, or | |
1503 | * iterate_devices() itself internally. | |
1504 | * | |
1505 | * For some target type (e.g. dm-stripe), one call of iterate_devices() may | |
1506 | * iterate multiple underlying devices internally, in which case a non-zero | |
1507 | * return code returned by iterate_devices_callout_fn will stop the iteration | |
1508 | * in advance. | |
1509 | * | |
1510 | * Cases requiring _any_ underlying device supporting some kind of attribute, | |
1511 | * should use the iteration structure like dm_table_any_dev_attr(), or call | |
1512 | * it directly. @func should handle semantics of positive examples, e.g. | |
1513 | * capable of something. | |
1514 | * | |
1515 | * Cases requiring _all_ underlying devices supporting some kind of attribute, | |
1516 | * should use the iteration structure like dm_table_supports_nowait() or | |
1517 | * dm_table_supports_discards(). Or introduce dm_table_all_devs_attr() that | |
1518 | * uses an @anti_func that handle semantics of counter examples, e.g. not | |
24f6b603 | 1519 | * capable of something. So: return !dm_table_any_dev_attr(t, anti_func, data); |
a4c8dd9c JX |
1520 | */ |
1521 | static bool dm_table_any_dev_attr(struct dm_table *t, | |
24f6b603 | 1522 | iterate_devices_callout_fn func, void *data) |
a4c8dd9c | 1523 | { |
564b5c54 MS |
1524 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1525 | struct dm_target *ti = dm_table_get_target(t, i); | |
a4c8dd9c JX |
1526 | |
1527 | if (ti->type->iterate_devices && | |
24f6b603 | 1528 | ti->type->iterate_devices(ti, func, data)) |
a4c8dd9c | 1529 | return true; |
255e2646 | 1530 | } |
a4c8dd9c JX |
1531 | |
1532 | return false; | |
1533 | } | |
1534 | ||
3ae70656 MS |
1535 | static int count_device(struct dm_target *ti, struct dm_dev *dev, |
1536 | sector_t start, sector_t len, void *data) | |
1537 | { | |
86a3238c | 1538 | unsigned int *num_devices = data; |
3ae70656 MS |
1539 | |
1540 | (*num_devices)++; | |
1541 | ||
1542 | return 0; | |
1543 | } | |
1544 | ||
9571f829 | 1545 | static bool dm_table_supports_poll(struct dm_table *t) |
b99fdcdc | 1546 | { |
564b5c54 MS |
1547 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1548 | struct dm_target *ti = dm_table_get_target(t, i); | |
9571f829 MS |
1549 | |
1550 | if (!ti->type->iterate_devices || | |
1551 | ti->type->iterate_devices(ti, device_not_poll_capable, NULL)) | |
1552 | return false; | |
1553 | } | |
1554 | ||
1555 | return true; | |
b99fdcdc ML |
1556 | } |
1557 | ||
3ae70656 MS |
1558 | /* |
1559 | * Check whether a table has no data devices attached using each | |
1560 | * target's iterate_devices method. | |
1561 | * Returns false if the result is unknown because a target doesn't | |
1562 | * support iterate_devices. | |
1563 | */ | |
564b5c54 | 1564 | bool dm_table_has_no_data_devices(struct dm_table *t) |
3ae70656 | 1565 | { |
564b5c54 MS |
1566 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1567 | struct dm_target *ti = dm_table_get_target(t, i); | |
86a3238c | 1568 | unsigned int num_devices = 0; |
3ae70656 MS |
1569 | |
1570 | if (!ti->type->iterate_devices) | |
1571 | return false; | |
1572 | ||
1573 | ti->type->iterate_devices(ti, count_device, &num_devices); | |
1574 | if (num_devices) | |
1575 | return false; | |
1576 | } | |
1577 | ||
1578 | return true; | |
1579 | } | |
1580 | ||
24f6b603 JX |
1581 | static int device_not_zoned_model(struct dm_target *ti, struct dm_dev *dev, |
1582 | sector_t start, sector_t len, void *data) | |
dd88d313 DLM |
1583 | { |
1584 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1585 | enum blk_zoned_model *zoned_model = data; | |
1586 | ||
cccb493c | 1587 | return blk_queue_zoned_model(q) != *zoned_model; |
dd88d313 DLM |
1588 | } |
1589 | ||
2d669ceb SK |
1590 | /* |
1591 | * Check the device zoned model based on the target feature flag. If the target | |
1592 | * has the DM_TARGET_ZONED_HM feature flag set, host-managed zoned devices are | |
1593 | * also accepted but all devices must have the same zoned model. If the target | |
1594 | * has the DM_TARGET_MIXED_ZONED_MODEL feature set, the devices can have any | |
1595 | * zoned model with all zoned devices having the same zone size. | |
1596 | */ | |
dd88d313 DLM |
1597 | static bool dm_table_supports_zoned_model(struct dm_table *t, |
1598 | enum blk_zoned_model zoned_model) | |
1599 | { | |
564b5c54 MS |
1600 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1601 | struct dm_target *ti = dm_table_get_target(t, i); | |
dd88d313 | 1602 | |
2d669ceb SK |
1603 | if (dm_target_supports_zoned_hm(ti->type)) { |
1604 | if (!ti->type->iterate_devices || | |
1605 | ti->type->iterate_devices(ti, device_not_zoned_model, | |
1606 | &zoned_model)) | |
1607 | return false; | |
1608 | } else if (!dm_target_supports_mixed_zoned_model(ti->type)) { | |
1609 | if (zoned_model == BLK_ZONED_HM) | |
1610 | return false; | |
1611 | } | |
dd88d313 DLM |
1612 | } |
1613 | ||
1614 | return true; | |
1615 | } | |
1616 | ||
24f6b603 JX |
1617 | static int device_not_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev, |
1618 | sector_t start, sector_t len, void *data) | |
dd88d313 | 1619 | { |
dd88d313 DLM |
1620 | unsigned int *zone_sectors = data; |
1621 | ||
edd1dbc8 | 1622 | if (!bdev_is_zoned(dev->bdev)) |
2d669ceb | 1623 | return 0; |
de71973c | 1624 | return bdev_zone_sectors(dev->bdev) != *zone_sectors; |
dd88d313 DLM |
1625 | } |
1626 | ||
2d669ceb SK |
1627 | /* |
1628 | * Check consistency of zoned model and zone sectors across all targets. For | |
1629 | * zone sectors, if the destination device is a zoned block device, it shall | |
1630 | * have the specified zone_sectors. | |
1631 | */ | |
564b5c54 | 1632 | static int validate_hardware_zoned_model(struct dm_table *t, |
dd88d313 DLM |
1633 | enum blk_zoned_model zoned_model, |
1634 | unsigned int zone_sectors) | |
1635 | { | |
1636 | if (zoned_model == BLK_ZONED_NONE) | |
1637 | return 0; | |
1638 | ||
564b5c54 | 1639 | if (!dm_table_supports_zoned_model(t, zoned_model)) { |
dd88d313 | 1640 | DMERR("%s: zoned model is not consistent across all devices", |
564b5c54 | 1641 | dm_device_name(t->md)); |
dd88d313 DLM |
1642 | return -EINVAL; |
1643 | } | |
1644 | ||
1645 | /* Check zone size validity and compatibility */ | |
1646 | if (!zone_sectors || !is_power_of_2(zone_sectors)) | |
1647 | return -EINVAL; | |
1648 | ||
564b5c54 | 1649 | if (dm_table_any_dev_attr(t, device_not_matches_zone_sectors, &zone_sectors)) { |
2d669ceb | 1650 | DMERR("%s: zone sectors is not consistent across all zoned devices", |
564b5c54 | 1651 | dm_device_name(t->md)); |
dd88d313 DLM |
1652 | return -EINVAL; |
1653 | } | |
1654 | ||
1655 | return 0; | |
1656 | } | |
1657 | ||
754c5fc7 MS |
1658 | /* |
1659 | * Establish the new table's queue_limits and validate them. | |
1660 | */ | |
564b5c54 | 1661 | int dm_calculate_queue_limits(struct dm_table *t, |
754c5fc7 MS |
1662 | struct queue_limits *limits) |
1663 | { | |
754c5fc7 | 1664 | struct queue_limits ti_limits; |
dd88d313 DLM |
1665 | enum blk_zoned_model zoned_model = BLK_ZONED_NONE; |
1666 | unsigned int zone_sectors = 0; | |
754c5fc7 | 1667 | |
b1bd055d | 1668 | blk_set_stacking_limits(limits); |
754c5fc7 | 1669 | |
564b5c54 MS |
1670 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1671 | struct dm_target *ti = dm_table_get_target(t, i); | |
754c5fc7 | 1672 | |
564b5c54 | 1673 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 | 1674 | |
85c938e8 MP |
1675 | if (!ti->type->iterate_devices) { |
1676 | /* Set I/O hints portion of queue limits */ | |
1677 | if (ti->type->io_hints) | |
1678 | ti->type->io_hints(ti, &ti_limits); | |
754c5fc7 | 1679 | goto combine_limits; |
85c938e8 | 1680 | } |
754c5fc7 MS |
1681 | |
1682 | /* | |
1683 | * Combine queue limits of all the devices this target uses. | |
1684 | */ | |
1685 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1686 | &ti_limits); | |
1687 | ||
dd88d313 DLM |
1688 | if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) { |
1689 | /* | |
1690 | * After stacking all limits, validate all devices | |
1691 | * in table support this zoned model and zone sectors. | |
1692 | */ | |
1693 | zoned_model = ti_limits.zoned; | |
1694 | zone_sectors = ti_limits.chunk_sectors; | |
1695 | } | |
1696 | ||
40bea431 MS |
1697 | /* Set I/O hints portion of queue limits */ |
1698 | if (ti->type->io_hints) | |
1699 | ti->type->io_hints(ti, &ti_limits); | |
1700 | ||
754c5fc7 MS |
1701 | /* |
1702 | * Check each device area is consistent with the target's | |
1703 | * overall queue limits. | |
1704 | */ | |
f6a1ed10 MP |
1705 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1706 | &ti_limits)) | |
754c5fc7 MS |
1707 | return -EINVAL; |
1708 | ||
1709 | combine_limits: | |
1710 | /* | |
1711 | * Merge this target's queue limits into the overall limits | |
1712 | * for the table. | |
1713 | */ | |
1714 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
2e84fecf | 1715 | DMWARN("%s: adding target device (start sect %llu len %llu) " |
b27d7f16 | 1716 | "caused an alignment inconsistency", |
564b5c54 | 1717 | dm_device_name(t->md), |
754c5fc7 MS |
1718 | (unsigned long long) ti->begin, |
1719 | (unsigned long long) ti->len); | |
1720 | } | |
1721 | ||
dd88d313 DLM |
1722 | /* |
1723 | * Verify that the zoned model and zone sectors, as determined before | |
1724 | * any .io_hints override, are the same across all devices in the table. | |
1725 | * - this is especially relevant if .io_hints is emulating a disk-managed | |
1726 | * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices. | |
1727 | * BUT... | |
1728 | */ | |
1729 | if (limits->zoned != BLK_ZONED_NONE) { | |
1730 | /* | |
1731 | * ...IF the above limits stacking determined a zoned model | |
1732 | * validate that all of the table's devices conform to it. | |
1733 | */ | |
1734 | zoned_model = limits->zoned; | |
1735 | zone_sectors = limits->chunk_sectors; | |
1736 | } | |
564b5c54 | 1737 | if (validate_hardware_zoned_model(t, zoned_model, zone_sectors)) |
dd88d313 DLM |
1738 | return -EINVAL; |
1739 | ||
564b5c54 | 1740 | return validate_hardware_logical_block_alignment(t, limits); |
754c5fc7 MS |
1741 | } |
1742 | ||
9c47008d | 1743 | /* |
25520d55 MP |
1744 | * Verify that all devices have an integrity profile that matches the |
1745 | * DM device's registered integrity profile. If the profiles don't | |
1746 | * match then unregister the DM device's integrity profile. | |
9c47008d | 1747 | */ |
25520d55 | 1748 | static void dm_table_verify_integrity(struct dm_table *t) |
9c47008d | 1749 | { |
a63a5cf8 | 1750 | struct gendisk *template_disk = NULL; |
9c47008d | 1751 | |
9b4b5a79 MB |
1752 | if (t->integrity_added) |
1753 | return; | |
1754 | ||
25520d55 MP |
1755 | if (t->integrity_supported) { |
1756 | /* | |
1757 | * Verify that the original integrity profile | |
1758 | * matches all the devices in this table. | |
1759 | */ | |
1760 | template_disk = dm_table_get_integrity_disk(t); | |
1761 | if (template_disk && | |
1762 | blk_integrity_compare(dm_disk(t->md), template_disk) >= 0) | |
1763 | return; | |
1764 | } | |
9c47008d | 1765 | |
25520d55 | 1766 | if (integrity_profile_exists(dm_disk(t->md))) { |
876fbba1 MS |
1767 | DMWARN("%s: unable to establish an integrity profile", |
1768 | dm_device_name(t->md)); | |
25520d55 MP |
1769 | blk_integrity_unregister(dm_disk(t->md)); |
1770 | } | |
9c47008d MP |
1771 | } |
1772 | ||
ed8b752b MS |
1773 | static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, |
1774 | sector_t start, sector_t len, void *data) | |
1775 | { | |
c888a8f9 | 1776 | unsigned long flush = (unsigned long) data; |
ed8b752b MS |
1777 | struct request_queue *q = bdev_get_queue(dev->bdev); |
1778 | ||
cccb493c | 1779 | return (q->queue_flags & flush); |
ed8b752b MS |
1780 | } |
1781 | ||
c888a8f9 | 1782 | static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush) |
ed8b752b | 1783 | { |
ed8b752b MS |
1784 | /* |
1785 | * Require at least one underlying device to support flushes. | |
1786 | * t->devices includes internal dm devices such as mirror logs | |
1787 | * so we need to use iterate_devices here, which targets | |
1788 | * supporting flushes must provide. | |
1789 | */ | |
564b5c54 MS |
1790 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1791 | struct dm_target *ti = dm_table_get_target(t, i); | |
ed8b752b | 1792 | |
55a62eef | 1793 | if (!ti->num_flush_bios) |
ed8b752b MS |
1794 | continue; |
1795 | ||
0e9c24ed | 1796 | if (ti->flush_supported) |
7f61f5a0 | 1797 | return true; |
0e9c24ed | 1798 | |
ed8b752b | 1799 | if (ti->type->iterate_devices && |
c888a8f9 | 1800 | ti->type->iterate_devices(ti, device_flush_capable, (void *) flush)) |
7f61f5a0 | 1801 | return true; |
ed8b752b MS |
1802 | } |
1803 | ||
7f61f5a0 | 1804 | return false; |
ed8b752b MS |
1805 | } |
1806 | ||
273752c9 VG |
1807 | static int device_dax_write_cache_enabled(struct dm_target *ti, |
1808 | struct dm_dev *dev, sector_t start, | |
1809 | sector_t len, void *data) | |
1810 | { | |
1811 | struct dax_device *dax_dev = dev->dax_dev; | |
1812 | ||
1813 | if (!dax_dev) | |
1814 | return false; | |
1815 | ||
1816 | if (dax_write_cache_enabled(dax_dev)) | |
1817 | return true; | |
1818 | return false; | |
1819 | } | |
1820 | ||
a4c8dd9c JX |
1821 | static int device_is_rotational(struct dm_target *ti, struct dm_dev *dev, |
1822 | sector_t start, sector_t len, void *data) | |
4693c966 | 1823 | { |
10f0d2a5 | 1824 | return !bdev_nonrot(dev->bdev); |
4693c966 MSB |
1825 | } |
1826 | ||
c3c4555e MB |
1827 | static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, |
1828 | sector_t start, sector_t len, void *data) | |
1829 | { | |
1830 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1831 | ||
cccb493c | 1832 | return !blk_queue_add_random(q); |
c3c4555e MB |
1833 | } |
1834 | ||
ac62d620 CH |
1835 | static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev, |
1836 | sector_t start, sector_t len, void *data) | |
1837 | { | |
1838 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1839 | ||
cccb493c | 1840 | return !q->limits.max_write_zeroes_sectors; |
ac62d620 CH |
1841 | } |
1842 | ||
1843 | static bool dm_table_supports_write_zeroes(struct dm_table *t) | |
1844 | { | |
564b5c54 MS |
1845 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1846 | struct dm_target *ti = dm_table_get_target(t, i); | |
ac62d620 CH |
1847 | |
1848 | if (!ti->num_write_zeroes_bios) | |
1849 | return false; | |
1850 | ||
1851 | if (!ti->type->iterate_devices || | |
1852 | ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL)) | |
1853 | return false; | |
1854 | } | |
1855 | ||
1856 | return true; | |
1857 | } | |
1858 | ||
6abc4946 KK |
1859 | static int device_not_nowait_capable(struct dm_target *ti, struct dm_dev *dev, |
1860 | sector_t start, sector_t len, void *data) | |
1861 | { | |
568ec936 | 1862 | return !bdev_nowait(dev->bdev); |
6abc4946 KK |
1863 | } |
1864 | ||
1865 | static bool dm_table_supports_nowait(struct dm_table *t) | |
1866 | { | |
564b5c54 MS |
1867 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1868 | struct dm_target *ti = dm_table_get_target(t, i); | |
6abc4946 KK |
1869 | |
1870 | if (!dm_target_supports_nowait(ti->type)) | |
1871 | return false; | |
1872 | ||
1873 | if (!ti->type->iterate_devices || | |
1874 | ti->type->iterate_devices(ti, device_not_nowait_capable, NULL)) | |
1875 | return false; | |
1876 | } | |
1877 | ||
1878 | return true; | |
1879 | } | |
1880 | ||
8a74d29d MS |
1881 | static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1882 | sector_t start, sector_t len, void *data) | |
a7ffb6a5 | 1883 | { |
70200574 | 1884 | return !bdev_max_discard_sectors(dev->bdev); |
a7ffb6a5 MP |
1885 | } |
1886 | ||
1887 | static bool dm_table_supports_discards(struct dm_table *t) | |
1888 | { | |
564b5c54 MS |
1889 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1890 | struct dm_target *ti = dm_table_get_target(t, i); | |
a7ffb6a5 MP |
1891 | |
1892 | if (!ti->num_discard_bios) | |
8a74d29d | 1893 | return false; |
a7ffb6a5 | 1894 | |
8a74d29d MS |
1895 | /* |
1896 | * Either the target provides discard support (as implied by setting | |
1897 | * 'discards_supported') or it relies on _all_ data devices having | |
1898 | * discard support. | |
1899 | */ | |
1900 | if (!ti->discards_supported && | |
1901 | (!ti->type->iterate_devices || | |
1902 | ti->type->iterate_devices(ti, device_not_discard_capable, NULL))) | |
1903 | return false; | |
a7ffb6a5 MP |
1904 | } |
1905 | ||
8a74d29d | 1906 | return true; |
a7ffb6a5 MP |
1907 | } |
1908 | ||
00716545 DS |
1909 | static int device_not_secure_erase_capable(struct dm_target *ti, |
1910 | struct dm_dev *dev, sector_t start, | |
1911 | sector_t len, void *data) | |
1912 | { | |
44abff2c | 1913 | return !bdev_max_secure_erase_sectors(dev->bdev); |
00716545 DS |
1914 | } |
1915 | ||
1916 | static bool dm_table_supports_secure_erase(struct dm_table *t) | |
1917 | { | |
564b5c54 MS |
1918 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1919 | struct dm_target *ti = dm_table_get_target(t, i); | |
00716545 DS |
1920 | |
1921 | if (!ti->num_secure_erase_bios) | |
1922 | return false; | |
1923 | ||
1924 | if (!ti->type->iterate_devices || | |
1925 | ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL)) | |
1926 | return false; | |
1927 | } | |
1928 | ||
1929 | return true; | |
1930 | } | |
1931 | ||
eb40c0ac ID |
1932 | static int device_requires_stable_pages(struct dm_target *ti, |
1933 | struct dm_dev *dev, sector_t start, | |
1934 | sector_t len, void *data) | |
1935 | { | |
36d25489 | 1936 | return bdev_stable_writes(dev->bdev); |
eb40c0ac ID |
1937 | } |
1938 | ||
bb37d772 DLM |
1939 | int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1940 | struct queue_limits *limits) | |
1da177e4 | 1941 | { |
519a7e16 | 1942 | bool wc = false, fua = false; |
bb37d772 | 1943 | int r; |
ed8b752b | 1944 | |
1da177e4 | 1945 | /* |
1197764e | 1946 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1947 | */ |
754c5fc7 | 1948 | q->limits = *limits; |
c9a3f6d6 | 1949 | |
6abc4946 KK |
1950 | if (dm_table_supports_nowait(t)) |
1951 | blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q); | |
1952 | else | |
1953 | blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, q); | |
1954 | ||
5d47c89f | 1955 | if (!dm_table_supports_discards(t)) { |
5d47c89f MS |
1956 | q->limits.max_discard_sectors = 0; |
1957 | q->limits.max_hw_discard_sectors = 0; | |
1958 | q->limits.discard_granularity = 0; | |
1959 | q->limits.discard_alignment = 0; | |
1960 | q->limits.discard_misaligned = 0; | |
70200574 | 1961 | } |
5ae89a87 | 1962 | |
44abff2c CH |
1963 | if (!dm_table_supports_secure_erase(t)) |
1964 | q->limits.max_secure_erase_sectors = 0; | |
00716545 | 1965 | |
c888a8f9 | 1966 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) { |
519a7e16 | 1967 | wc = true; |
c888a8f9 | 1968 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA))) |
519a7e16 | 1969 | fua = true; |
ed8b752b | 1970 | } |
519a7e16 | 1971 | blk_queue_write_cache(q, wc, fua); |
ed8b752b | 1972 | |
7b0800d0 | 1973 | if (dm_table_supports_dax(t, device_not_dax_capable)) { |
8b904b5b | 1974 | blk_queue_flag_set(QUEUE_FLAG_DAX, q); |
7b0800d0 | 1975 | if (dm_table_supports_dax(t, device_not_dax_synchronous_capable)) |
2e9ee095 | 1976 | set_dax_synchronous(t->md->dax_dev); |
03b18887 | 1977 | } else |
dbc62659 RZ |
1978 | blk_queue_flag_clear(QUEUE_FLAG_DAX, q); |
1979 | ||
24f6b603 | 1980 | if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL)) |
273752c9 VG |
1981 | dax_write_cache(t->md->dax_dev, true); |
1982 | ||
c3c4555e | 1983 | /* Ensure that all underlying devices are non-rotational. */ |
24f6b603 | 1984 | if (dm_table_any_dev_attr(t, device_is_rotational, NULL)) |
8b904b5b | 1985 | blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); |
a4c8dd9c JX |
1986 | else |
1987 | blk_queue_flag_set(QUEUE_FLAG_NONROT, q); | |
4693c966 | 1988 | |
ac62d620 CH |
1989 | if (!dm_table_supports_write_zeroes(t)) |
1990 | q->limits.max_write_zeroes_sectors = 0; | |
c1a94672 | 1991 | |
25520d55 | 1992 | dm_table_verify_integrity(t); |
e6ee8c0b | 1993 | |
eb40c0ac ID |
1994 | /* |
1995 | * Some devices don't use blk_integrity but still want stable pages | |
1996 | * because they do their own checksumming. | |
a4c8dd9c JX |
1997 | * If any underlying device requires stable pages, a table must require |
1998 | * them as well. Only targets that support iterate_devices are considered: | |
1999 | * don't want error, zero, etc to require stable pages. | |
eb40c0ac | 2000 | */ |
24f6b603 | 2001 | if (dm_table_any_dev_attr(t, device_requires_stable_pages, NULL)) |
1cb039f3 | 2002 | blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); |
eb40c0ac | 2003 | else |
1cb039f3 | 2004 | blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q); |
eb40c0ac | 2005 | |
c3c4555e MB |
2006 | /* |
2007 | * Determine whether or not this queue's I/O timings contribute | |
2008 | * to the entropy pool, Only request-based targets use this. | |
2009 | * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not | |
2010 | * have it set. | |
2011 | */ | |
24f6b603 JX |
2012 | if (blk_queue_add_random(q) && |
2013 | dm_table_any_dev_attr(t, device_is_not_random, NULL)) | |
8b904b5b | 2014 | blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); |
bf505456 | 2015 | |
bb37d772 DLM |
2016 | /* |
2017 | * For a zoned target, setup the zones related queue attributes | |
2018 | * and resources necessary for zone append emulation if necessary. | |
2019 | */ | |
2020 | if (blk_queue_is_zoned(q)) { | |
2021 | r = dm_set_zones_restrictions(t, q); | |
2022 | if (r) | |
2023 | return r; | |
442761fd MS |
2024 | if (!static_key_enabled(&zoned_enabled.key)) |
2025 | static_branch_enable(&zoned_enabled); | |
bb37d772 | 2026 | } |
c6d6e9b0 | 2027 | |
cb77cb5a | 2028 | dm_update_crypto_profile(q, t); |
471aa704 | 2029 | disk_update_readahead(t->md->disk); |
bb37d772 | 2030 | |
b99fdcdc ML |
2031 | /* |
2032 | * Check for request-based device is left to | |
2033 | * dm_mq_init_request_queue()->blk_mq_init_allocated_queue(). | |
2034 | * | |
2035 | * For bio-based device, only set QUEUE_FLAG_POLL when all | |
2036 | * underlying devices supporting polling. | |
2037 | */ | |
2038 | if (__table_type_bio_based(t->type)) { | |
2039 | if (dm_table_supports_poll(t)) | |
2040 | blk_queue_flag_set(QUEUE_FLAG_POLL, q); | |
2041 | else | |
2042 | blk_queue_flag_clear(QUEUE_FLAG_POLL, q); | |
2043 | } | |
2044 | ||
bb37d772 | 2045 | return 0; |
1da177e4 LT |
2046 | } |
2047 | ||
1da177e4 LT |
2048 | struct list_head *dm_table_get_devices(struct dm_table *t) |
2049 | { | |
2050 | return &t->devices; | |
2051 | } | |
2052 | ||
05bdb996 | 2053 | blk_mode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
2054 | { |
2055 | return t->mode; | |
2056 | } | |
08649012 | 2057 | EXPORT_SYMBOL(dm_table_get_mode); |
1da177e4 | 2058 | |
d67ee213 MS |
2059 | enum suspend_mode { |
2060 | PRESUSPEND, | |
2061 | PRESUSPEND_UNDO, | |
2062 | POSTSUSPEND, | |
2063 | }; | |
2064 | ||
2065 | static void suspend_targets(struct dm_table *t, enum suspend_mode mode) | |
1da177e4 | 2066 | { |
1ea0654e BVA |
2067 | lockdep_assert_held(&t->md->suspend_lock); |
2068 | ||
564b5c54 MS |
2069 | for (unsigned int i = 0; i < t->num_targets; i++) { |
2070 | struct dm_target *ti = dm_table_get_target(t, i); | |
2071 | ||
d67ee213 MS |
2072 | switch (mode) { |
2073 | case PRESUSPEND: | |
2074 | if (ti->type->presuspend) | |
2075 | ti->type->presuspend(ti); | |
2076 | break; | |
2077 | case PRESUSPEND_UNDO: | |
2078 | if (ti->type->presuspend_undo) | |
2079 | ti->type->presuspend_undo(ti); | |
2080 | break; | |
2081 | case POSTSUSPEND: | |
1da177e4 LT |
2082 | if (ti->type->postsuspend) |
2083 | ti->type->postsuspend(ti); | |
d67ee213 MS |
2084 | break; |
2085 | } | |
1da177e4 LT |
2086 | } |
2087 | } | |
2088 | ||
2089 | void dm_table_presuspend_targets(struct dm_table *t) | |
2090 | { | |
cf222b37 AK |
2091 | if (!t) |
2092 | return; | |
2093 | ||
d67ee213 MS |
2094 | suspend_targets(t, PRESUSPEND); |
2095 | } | |
2096 | ||
2097 | void dm_table_presuspend_undo_targets(struct dm_table *t) | |
2098 | { | |
2099 | if (!t) | |
2100 | return; | |
2101 | ||
2102 | suspend_targets(t, PRESUSPEND_UNDO); | |
1da177e4 LT |
2103 | } |
2104 | ||
2105 | void dm_table_postsuspend_targets(struct dm_table *t) | |
2106 | { | |
cf222b37 AK |
2107 | if (!t) |
2108 | return; | |
2109 | ||
d67ee213 | 2110 | suspend_targets(t, POSTSUSPEND); |
1da177e4 LT |
2111 | } |
2112 | ||
8757b776 | 2113 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 2114 | { |
564b5c54 MS |
2115 | unsigned int i; |
2116 | int r = 0; | |
8757b776 | 2117 | |
1ea0654e BVA |
2118 | lockdep_assert_held(&t->md->suspend_lock); |
2119 | ||
8757b776 | 2120 | for (i = 0; i < t->num_targets; i++) { |
564b5c54 | 2121 | struct dm_target *ti = dm_table_get_target(t, i); |
8757b776 MB |
2122 | |
2123 | if (!ti->type->preresume) | |
2124 | continue; | |
2125 | ||
2126 | r = ti->type->preresume(ti); | |
7833b08e MS |
2127 | if (r) { |
2128 | DMERR("%s: %s: preresume failed, error = %d", | |
2129 | dm_device_name(t->md), ti->type->name, r); | |
8757b776 | 2130 | return r; |
7833b08e | 2131 | } |
8757b776 | 2132 | } |
1da177e4 LT |
2133 | |
2134 | for (i = 0; i < t->num_targets; i++) { | |
564b5c54 | 2135 | struct dm_target *ti = dm_table_get_target(t, i); |
1da177e4 LT |
2136 | |
2137 | if (ti->type->resume) | |
2138 | ti->type->resume(ti); | |
2139 | } | |
8757b776 MB |
2140 | |
2141 | return 0; | |
1da177e4 LT |
2142 | } |
2143 | ||
1134e5ae MA |
2144 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
2145 | { | |
1134e5ae MA |
2146 | return t->md; |
2147 | } | |
08649012 | 2148 | EXPORT_SYMBOL(dm_table_get_md); |
1134e5ae | 2149 | |
f349b0a3 MM |
2150 | const char *dm_table_device_name(struct dm_table *t) |
2151 | { | |
2152 | return dm_device_name(t->md); | |
2153 | } | |
2154 | EXPORT_SYMBOL_GPL(dm_table_device_name); | |
2155 | ||
9974fa2c MS |
2156 | void dm_table_run_md_queue_async(struct dm_table *t) |
2157 | { | |
9974fa2c MS |
2158 | if (!dm_table_request_based(t)) |
2159 | return; | |
2160 | ||
33bd6f06 MS |
2161 | if (t->md->queue) |
2162 | blk_mq_run_hw_queues(t->md->queue, true); | |
9974fa2c MS |
2163 | } |
2164 | EXPORT_SYMBOL(dm_table_run_md_queue_async); | |
2165 |