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 | ||
aeb5d727 | 129 | int dm_table_create(struct dm_table **result, fmode_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); |
1da177e4 LT |
138 | |
139 | if (!num_targets) | |
140 | num_targets = KEYS_PER_NODE; | |
141 | ||
142 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
143 | ||
5b2d0657 MP |
144 | if (!num_targets) { |
145 | kfree(t); | |
146 | return -ENOMEM; | |
147 | } | |
148 | ||
1da177e4 LT |
149 | if (alloc_targets(t, num_targets)) { |
150 | kfree(t); | |
1da177e4 LT |
151 | return -ENOMEM; |
152 | } | |
153 | ||
e83068a5 | 154 | t->type = DM_TYPE_NONE; |
1da177e4 | 155 | t->mode = mode; |
1134e5ae | 156 | t->md = md; |
1da177e4 LT |
157 | *result = t; |
158 | return 0; | |
159 | } | |
160 | ||
86f1152b | 161 | static void free_devices(struct list_head *devices, struct mapped_device *md) |
1da177e4 LT |
162 | { |
163 | struct list_head *tmp, *next; | |
164 | ||
afb24528 | 165 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
166 | struct dm_dev_internal *dd = |
167 | list_entry(tmp, struct dm_dev_internal, list); | |
86f1152b BM |
168 | DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s", |
169 | dm_device_name(md), dd->dm_dev->name); | |
170 | dm_put_table_device(md, dd->dm_dev); | |
1da177e4 LT |
171 | kfree(dd); |
172 | } | |
173 | } | |
174 | ||
cb77cb5a | 175 | static void dm_table_destroy_crypto_profile(struct dm_table *t); |
aa6ce87a | 176 | |
d5816876 | 177 | void dm_table_destroy(struct dm_table *t) |
1da177e4 | 178 | { |
a7940155 AK |
179 | if (!t) |
180 | return; | |
181 | ||
26803b9f | 182 | /* free the indexes */ |
1da177e4 | 183 | if (t->depth >= 2) |
7a35693a | 184 | kvfree(t->index[t->depth - 2]); |
1da177e4 LT |
185 | |
186 | /* free the targets */ | |
564b5c54 MS |
187 | for (unsigned int i = 0; i < t->num_targets; i++) { |
188 | struct dm_target *ti = dm_table_get_target(t, i); | |
1da177e4 | 189 | |
564b5c54 MS |
190 | if (ti->type->dtr) |
191 | ti->type->dtr(ti); | |
1da177e4 | 192 | |
564b5c54 | 193 | dm_put_target_type(ti->type); |
1da177e4 LT |
194 | } |
195 | ||
7a35693a | 196 | kvfree(t->highs); |
1da177e4 LT |
197 | |
198 | /* free the device list */ | |
86f1152b | 199 | free_devices(&t->devices, t->md); |
1da177e4 | 200 | |
e6ee8c0b KU |
201 | dm_free_md_mempools(t->mempools); |
202 | ||
cb77cb5a | 203 | dm_table_destroy_crypto_profile(t); |
aa6ce87a | 204 | |
1da177e4 LT |
205 | kfree(t); |
206 | } | |
207 | ||
1da177e4 LT |
208 | /* |
209 | * See if we've already got a device in the list. | |
210 | */ | |
82b1519b | 211 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 212 | { |
82b1519b | 213 | struct dm_dev_internal *dd; |
1da177e4 | 214 | |
43be9c74 | 215 | list_for_each_entry(dd, l, list) |
86f1152b | 216 | if (dd->dm_dev->bdev->bd_dev == dev) |
1da177e4 LT |
217 | return dd; |
218 | ||
219 | return NULL; | |
220 | } | |
221 | ||
1da177e4 | 222 | /* |
f6a1ed10 | 223 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 224 | */ |
f6a1ed10 MP |
225 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
226 | sector_t start, sector_t len, void *data) | |
1da177e4 | 227 | { |
754c5fc7 MS |
228 | struct queue_limits *limits = data; |
229 | struct block_device *bdev = dev->bdev; | |
6dcbb52c | 230 | sector_t dev_size = bdev_nr_sectors(bdev); |
02acc3a4 | 231 | unsigned short logical_block_size_sectors = |
754c5fc7 | 232 | limits->logical_block_size >> SECTOR_SHIFT; |
2cd54d9b MA |
233 | |
234 | if (!dev_size) | |
f6a1ed10 | 235 | return 0; |
2cd54d9b | 236 | |
5dea271b | 237 | if ((start >= dev_size) || (start + len > dev_size)) { |
2e84fecf | 238 | DMERR("%s: %pg too small for target: start=%llu, len=%llu, dev_size=%llu", |
43e6c111 MP |
239 | dm_device_name(ti->table->md), bdev, |
240 | (unsigned long long)start, | |
241 | (unsigned long long)len, | |
242 | (unsigned long long)dev_size); | |
f6a1ed10 | 243 | return 1; |
02acc3a4 MS |
244 | } |
245 | ||
dd88d313 DLM |
246 | /* |
247 | * If the target is mapped to zoned block device(s), check | |
248 | * that the zones are not partially mapped. | |
249 | */ | |
dd73c320 | 250 | if (bdev_is_zoned(bdev)) { |
dd88d313 DLM |
251 | unsigned int zone_sectors = bdev_zone_sectors(bdev); |
252 | ||
253 | if (start & (zone_sectors - 1)) { | |
43e6c111 MP |
254 | DMERR("%s: start=%llu not aligned to h/w zone size %u of %pg", |
255 | dm_device_name(ti->table->md), | |
256 | (unsigned long long)start, | |
257 | zone_sectors, bdev); | |
dd88d313 DLM |
258 | return 1; |
259 | } | |
260 | ||
261 | /* | |
262 | * Note: The last zone of a zoned block device may be smaller | |
263 | * than other zones. So for a target mapping the end of a | |
264 | * zoned block device with such a zone, len would not be zone | |
265 | * aligned. We do not allow such last smaller zone to be part | |
266 | * of the mapping here to ensure that mappings with multiple | |
267 | * devices do not end up with a smaller zone in the middle of | |
268 | * the sector range. | |
269 | */ | |
270 | if (len & (zone_sectors - 1)) { | |
43e6c111 MP |
271 | DMERR("%s: len=%llu not aligned to h/w zone size %u of %pg", |
272 | dm_device_name(ti->table->md), | |
273 | (unsigned long long)len, | |
274 | zone_sectors, bdev); | |
dd88d313 DLM |
275 | return 1; |
276 | } | |
277 | } | |
278 | ||
02acc3a4 | 279 | if (logical_block_size_sectors <= 1) |
f6a1ed10 | 280 | return 0; |
02acc3a4 MS |
281 | |
282 | if (start & (logical_block_size_sectors - 1)) { | |
2e84fecf | 283 | DMERR("%s: start=%llu not aligned to h/w logical block size %u of %pg", |
43e6c111 MP |
284 | dm_device_name(ti->table->md), |
285 | (unsigned long long)start, | |
286 | limits->logical_block_size, bdev); | |
f6a1ed10 | 287 | return 1; |
02acc3a4 MS |
288 | } |
289 | ||
5dea271b | 290 | if (len & (logical_block_size_sectors - 1)) { |
2e84fecf | 291 | DMERR("%s: len=%llu not aligned to h/w logical block size %u of %pg", |
43e6c111 MP |
292 | dm_device_name(ti->table->md), |
293 | (unsigned long long)len, | |
294 | limits->logical_block_size, bdev); | |
f6a1ed10 | 295 | return 1; |
02acc3a4 MS |
296 | } |
297 | ||
f6a1ed10 | 298 | return 0; |
1da177e4 LT |
299 | } |
300 | ||
301 | /* | |
570b9d96 | 302 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 303 | * careful to leave things as they were if we fail to reopen the |
570b9d96 | 304 | * device and not to touch the existing bdev field in case |
21cf8661 | 305 | * it is accessed concurrently. |
1da177e4 | 306 | */ |
aeb5d727 | 307 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 308 | struct mapped_device *md) |
1da177e4 LT |
309 | { |
310 | int r; | |
86f1152b | 311 | struct dm_dev *old_dev, *new_dev; |
1da177e4 | 312 | |
86f1152b | 313 | old_dev = dd->dm_dev; |
570b9d96 | 314 | |
86f1152b BM |
315 | r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, |
316 | dd->dm_dev->mode | new_mode, &new_dev); | |
570b9d96 AK |
317 | if (r) |
318 | return r; | |
1da177e4 | 319 | |
86f1152b BM |
320 | dd->dm_dev = new_dev; |
321 | dm_put_table_device(md, old_dev); | |
1da177e4 | 322 | |
570b9d96 | 323 | return 0; |
1da177e4 LT |
324 | } |
325 | ||
4df2bf46 D |
326 | /* |
327 | * Convert the path to a device | |
328 | */ | |
329 | dev_t dm_get_dev_t(const char *path) | |
330 | { | |
3c120169 | 331 | dev_t dev; |
4df2bf46 | 332 | |
4e7b5671 | 333 | if (lookup_bdev(path, &dev)) |
4df2bf46 | 334 | dev = name_to_dev_t(path); |
4df2bf46 D |
335 | return dev; |
336 | } | |
337 | EXPORT_SYMBOL_GPL(dm_get_dev_t); | |
338 | ||
1da177e4 LT |
339 | /* |
340 | * Add a device to the list, or just increment the usage count if | |
341 | * it's already present. | |
342 | */ | |
08649012 MS |
343 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
344 | struct dm_dev **result) | |
1da177e4 LT |
345 | { |
346 | int r; | |
4df2bf46 | 347 | dev_t dev; |
809b1e49 HR |
348 | unsigned int major, minor; |
349 | char dummy; | |
82b1519b | 350 | struct dm_dev_internal *dd; |
08649012 | 351 | struct dm_table *t = ti->table; |
1da177e4 | 352 | |
547bc926 | 353 | BUG_ON(!t); |
1da177e4 | 354 | |
809b1e49 HR |
355 | if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) { |
356 | /* Extract the major/minor numbers */ | |
357 | dev = MKDEV(major, minor); | |
358 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
359 | return -EOVERFLOW; | |
360 | } else { | |
361 | dev = dm_get_dev_t(path); | |
362 | if (!dev) | |
363 | return -ENODEV; | |
364 | } | |
d1c0e158 BM |
365 | if (dev == disk_devt(t->md->disk)) |
366 | return -EINVAL; | |
1da177e4 LT |
367 | |
368 | dd = find_device(&t->devices, dev); | |
369 | if (!dd) { | |
370 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
371 | if (!dd) | |
372 | return -ENOMEM; | |
373 | ||
d715fa23 HM |
374 | r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev); |
375 | if (r) { | |
1da177e4 LT |
376 | kfree(dd); |
377 | return r; | |
378 | } | |
379 | ||
2a0b4682 | 380 | refcount_set(&dd->count, 1); |
1da177e4 | 381 | list_add(&dd->list, &t->devices); |
afc567a4 | 382 | goto out; |
1da177e4 | 383 | |
86f1152b | 384 | } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { |
f165921d | 385 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
386 | if (r) |
387 | return r; | |
388 | } | |
afc567a4 MS |
389 | refcount_inc(&dd->count); |
390 | out: | |
86f1152b | 391 | *result = dd->dm_dev; |
1da177e4 LT |
392 | return 0; |
393 | } | |
08649012 | 394 | EXPORT_SYMBOL(dm_get_device); |
1da177e4 | 395 | |
11f0431b MS |
396 | static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
397 | sector_t start, sector_t len, void *data) | |
1da177e4 | 398 | { |
754c5fc7 MS |
399 | struct queue_limits *limits = data; |
400 | struct block_device *bdev = dev->bdev; | |
165125e1 | 401 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
402 | |
403 | if (unlikely(!q)) { | |
385411ff CH |
404 | DMWARN("%s: Cannot set limits for nonexistent device %pg", |
405 | dm_device_name(ti->table->md), bdev); | |
754c5fc7 | 406 | return 0; |
0c2322e4 | 407 | } |
3cb40214 | 408 | |
9efa82ef CH |
409 | if (blk_stack_limits(limits, &q->limits, |
410 | get_start_sect(bdev) + start) < 0) | |
385411ff | 411 | DMWARN("%s: adding target device %pg caused an alignment inconsistency: " |
a963a956 MS |
412 | "physical_block_size=%u, logical_block_size=%u, " |
413 | "alignment_offset=%u, start=%llu", | |
385411ff | 414 | dm_device_name(ti->table->md), bdev, |
a963a956 MS |
415 | q->limits.physical_block_size, |
416 | q->limits.logical_block_size, | |
417 | q->limits.alignment_offset, | |
b27d7f16 | 418 | (unsigned long long) start << SECTOR_SHIFT); |
754c5fc7 | 419 | return 0; |
3cb40214 | 420 | } |
969429b5 | 421 | |
1da177e4 | 422 | /* |
08649012 | 423 | * Decrement a device's use count and remove it if necessary. |
1da177e4 | 424 | */ |
82b1519b | 425 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 426 | { |
86f1152b BM |
427 | int found = 0; |
428 | struct list_head *devices = &ti->table->devices; | |
429 | struct dm_dev_internal *dd; | |
82b1519b | 430 | |
86f1152b BM |
431 | list_for_each_entry(dd, devices, list) { |
432 | if (dd->dm_dev == d) { | |
433 | found = 1; | |
434 | break; | |
435 | } | |
436 | } | |
437 | if (!found) { | |
43e6c111 MP |
438 | DMERR("%s: device %s not in table devices list", |
439 | dm_device_name(ti->table->md), d->name); | |
86f1152b BM |
440 | return; |
441 | } | |
2a0b4682 | 442 | if (refcount_dec_and_test(&dd->count)) { |
86f1152b | 443 | dm_put_table_device(ti->table->md, d); |
1da177e4 LT |
444 | list_del(&dd->list); |
445 | kfree(dd); | |
446 | } | |
447 | } | |
08649012 | 448 | EXPORT_SYMBOL(dm_put_device); |
1da177e4 LT |
449 | |
450 | /* | |
451 | * Checks to see if the target joins onto the end of the table. | |
452 | */ | |
564b5c54 | 453 | static int adjoin(struct dm_table *t, struct dm_target *ti) |
1da177e4 LT |
454 | { |
455 | struct dm_target *prev; | |
456 | ||
564b5c54 | 457 | if (!t->num_targets) |
1da177e4 LT |
458 | return !ti->begin; |
459 | ||
564b5c54 | 460 | prev = &t->targets[t->num_targets - 1]; |
1da177e4 LT |
461 | return (ti->begin == (prev->begin + prev->len)); |
462 | } | |
463 | ||
464 | /* | |
465 | * Used to dynamically allocate the arg array. | |
f36afb39 MP |
466 | * |
467 | * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must | |
468 | * process messages even if some device is suspended. These messages have a | |
469 | * small fixed number of arguments. | |
470 | * | |
471 | * On the other hand, dm-switch needs to process bulk data using messages and | |
472 | * excessive use of GFP_NOIO could cause trouble. | |
1da177e4 | 473 | */ |
86a3238c | 474 | static char **realloc_argv(unsigned int *size, char **old_argv) |
1da177e4 LT |
475 | { |
476 | char **argv; | |
86a3238c | 477 | unsigned int new_size; |
f36afb39 | 478 | gfp_t gfp; |
1da177e4 | 479 | |
610b15c5 KC |
480 | if (*size) { |
481 | new_size = *size * 2; | |
f36afb39 MP |
482 | gfp = GFP_KERNEL; |
483 | } else { | |
484 | new_size = 8; | |
485 | gfp = GFP_NOIO; | |
486 | } | |
6da2ec56 | 487 | argv = kmalloc_array(new_size, sizeof(*argv), gfp); |
a0651926 | 488 | if (argv && old_argv) { |
610b15c5 KC |
489 | memcpy(argv, old_argv, *size * sizeof(*argv)); |
490 | *size = new_size; | |
1da177e4 LT |
491 | } |
492 | ||
493 | kfree(old_argv); | |
494 | return argv; | |
495 | } | |
496 | ||
497 | /* | |
498 | * Destructively splits up the argument list to pass to ctr. | |
499 | */ | |
500 | int dm_split_args(int *argc, char ***argvp, char *input) | |
501 | { | |
502 | char *start, *end = input, *out, **argv = NULL; | |
86a3238c | 503 | unsigned int array_size = 0; |
1da177e4 LT |
504 | |
505 | *argc = 0; | |
814d6862 DT |
506 | |
507 | if (!input) { | |
508 | *argvp = NULL; | |
509 | return 0; | |
510 | } | |
511 | ||
1da177e4 LT |
512 | argv = realloc_argv(&array_size, argv); |
513 | if (!argv) | |
514 | return -ENOMEM; | |
515 | ||
516 | while (1) { | |
1da177e4 | 517 | /* Skip whitespace */ |
e7d2860b | 518 | start = skip_spaces(end); |
1da177e4 LT |
519 | |
520 | if (!*start) | |
521 | break; /* success, we hit the end */ | |
522 | ||
523 | /* 'out' is used to remove any back-quotes */ | |
524 | end = out = start; | |
525 | while (*end) { | |
526 | /* Everything apart from '\0' can be quoted */ | |
527 | if (*end == '\\' && *(end + 1)) { | |
528 | *out++ = *(end + 1); | |
529 | end += 2; | |
530 | continue; | |
531 | } | |
532 | ||
533 | if (isspace(*end)) | |
534 | break; /* end of token */ | |
535 | ||
536 | *out++ = *end++; | |
537 | } | |
538 | ||
539 | /* have we already filled the array ? */ | |
540 | if ((*argc + 1) > array_size) { | |
541 | argv = realloc_argv(&array_size, argv); | |
542 | if (!argv) | |
543 | return -ENOMEM; | |
544 | } | |
545 | ||
546 | /* we know this is whitespace */ | |
547 | if (*end) | |
548 | end++; | |
549 | ||
550 | /* terminate the string and put it in the array */ | |
551 | *out = '\0'; | |
552 | argv[*argc] = start; | |
553 | (*argc)++; | |
554 | } | |
555 | ||
556 | *argvp = argv; | |
557 | return 0; | |
558 | } | |
559 | ||
be6d4305 MS |
560 | /* |
561 | * Impose necessary and sufficient conditions on a devices's table such | |
562 | * that any incoming bio which respects its logical_block_size can be | |
563 | * processed successfully. If it falls across the boundary between | |
564 | * two or more targets, the size of each piece it gets split into must | |
565 | * be compatible with the logical_block_size of the target processing it. | |
566 | */ | |
564b5c54 MS |
567 | static int validate_hardware_logical_block_alignment(struct dm_table *t, |
568 | struct queue_limits *limits) | |
be6d4305 MS |
569 | { |
570 | /* | |
571 | * This function uses arithmetic modulo the logical_block_size | |
572 | * (in units of 512-byte sectors). | |
573 | */ | |
574 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 575 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
576 | |
577 | /* | |
578 | * Offset of the start of the next table entry, mod logical_block_size. | |
579 | */ | |
580 | unsigned short next_target_start = 0; | |
581 | ||
582 | /* | |
583 | * Given an aligned bio that extends beyond the end of a | |
584 | * target, how many sectors must the next target handle? | |
585 | */ | |
586 | unsigned short remaining = 0; | |
587 | ||
3f649ab7 | 588 | struct dm_target *ti; |
754c5fc7 | 589 | struct queue_limits ti_limits; |
564b5c54 | 590 | unsigned int i; |
be6d4305 MS |
591 | |
592 | /* | |
593 | * Check each entry in the table in turn. | |
594 | */ | |
564b5c54 MS |
595 | for (i = 0; i < t->num_targets; i++) { |
596 | ti = dm_table_get_target(t, i); | |
be6d4305 | 597 | |
b1bd055d | 598 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
599 | |
600 | /* combine all target devices' limits */ | |
601 | if (ti->type->iterate_devices) | |
602 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
603 | &ti_limits); | |
604 | ||
be6d4305 MS |
605 | /* |
606 | * If the remaining sectors fall entirely within this | |
607 | * table entry are they compatible with its logical_block_size? | |
608 | */ | |
609 | if (remaining < ti->len && | |
754c5fc7 | 610 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
611 | SECTOR_SHIFT) - 1)) |
612 | break; /* Error */ | |
613 | ||
614 | next_target_start = | |
615 | (unsigned short) ((next_target_start + ti->len) & | |
616 | (device_logical_block_size_sects - 1)); | |
617 | remaining = next_target_start ? | |
618 | device_logical_block_size_sects - next_target_start : 0; | |
619 | } | |
620 | ||
621 | if (remaining) { | |
43e6c111 MP |
622 | DMERR("%s: table line %u (start sect %llu len %llu) " |
623 | "not aligned to h/w logical block size %u", | |
624 | dm_device_name(t->md), i, | |
625 | (unsigned long long) ti->begin, | |
626 | (unsigned long long) ti->len, | |
627 | limits->logical_block_size); | |
be6d4305 MS |
628 | return -EINVAL; |
629 | } | |
630 | ||
631 | return 0; | |
632 | } | |
633 | ||
1da177e4 LT |
634 | int dm_table_add_target(struct dm_table *t, const char *type, |
635 | sector_t start, sector_t len, char *params) | |
636 | { | |
637 | int r = -EINVAL, argc; | |
638 | char **argv; | |
899ab445 | 639 | struct dm_target *ti; |
1da177e4 | 640 | |
3791e2fc AK |
641 | if (t->singleton) { |
642 | DMERR("%s: target type %s must appear alone in table", | |
643 | dm_device_name(t->md), t->targets->type->name); | |
644 | return -EINVAL; | |
645 | } | |
646 | ||
57a2f238 | 647 | BUG_ON(t->num_targets >= t->num_allocated); |
1da177e4 | 648 | |
899ab445 MS |
649 | ti = t->targets + t->num_targets; |
650 | memset(ti, 0, sizeof(*ti)); | |
1da177e4 LT |
651 | |
652 | if (!len) { | |
72d94861 | 653 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
654 | return -EINVAL; |
655 | } | |
656 | ||
899ab445 MS |
657 | ti->type = dm_get_target_type(type); |
658 | if (!ti->type) { | |
dafa724b | 659 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), type); |
1da177e4 LT |
660 | return -EINVAL; |
661 | } | |
662 | ||
899ab445 | 663 | if (dm_target_needs_singleton(ti->type)) { |
3791e2fc | 664 | if (t->num_targets) { |
899ab445 | 665 | ti->error = "singleton target type must appear alone in table"; |
dafa724b | 666 | goto bad; |
3791e2fc | 667 | } |
e83068a5 | 668 | t->singleton = true; |
3791e2fc AK |
669 | } |
670 | ||
899ab445 MS |
671 | if (dm_target_always_writeable(ti->type) && !(t->mode & FMODE_WRITE)) { |
672 | ti->error = "target type may not be included in a read-only table"; | |
dafa724b | 673 | goto bad; |
cc6cbe14 AK |
674 | } |
675 | ||
36a0456f | 676 | if (t->immutable_target_type) { |
899ab445 MS |
677 | if (t->immutable_target_type != ti->type) { |
678 | ti->error = "immutable target type cannot be mixed with other target types"; | |
dafa724b | 679 | goto bad; |
36a0456f | 680 | } |
899ab445 | 681 | } else if (dm_target_is_immutable(ti->type)) { |
36a0456f | 682 | if (t->num_targets) { |
899ab445 | 683 | ti->error = "immutable target type cannot be mixed with other target types"; |
dafa724b | 684 | goto bad; |
36a0456f | 685 | } |
899ab445 | 686 | t->immutable_target_type = ti->type; |
36a0456f AK |
687 | } |
688 | ||
899ab445 | 689 | if (dm_target_has_integrity(ti->type)) |
9b4b5a79 MB |
690 | t->integrity_added = 1; |
691 | ||
899ab445 MS |
692 | ti->table = t; |
693 | ti->begin = start; | |
694 | ti->len = len; | |
695 | ti->error = "Unknown error"; | |
1da177e4 LT |
696 | |
697 | /* | |
698 | * Does this target adjoin the previous one ? | |
699 | */ | |
899ab445 MS |
700 | if (!adjoin(t, ti)) { |
701 | ti->error = "Gap in table"; | |
1da177e4 LT |
702 | goto bad; |
703 | } | |
704 | ||
705 | r = dm_split_args(&argc, &argv, params); | |
706 | if (r) { | |
899ab445 | 707 | ti->error = "couldn't split parameters"; |
1da177e4 LT |
708 | goto bad; |
709 | } | |
710 | ||
899ab445 | 711 | r = ti->type->ctr(ti, argc, argv); |
1da177e4 LT |
712 | kfree(argv); |
713 | if (r) | |
714 | goto bad; | |
715 | ||
899ab445 | 716 | t->highs[t->num_targets++] = ti->begin + ti->len - 1; |
1da177e4 | 717 | |
899ab445 | 718 | if (!ti->num_discard_bios && ti->discards_supported) |
55a62eef | 719 | DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.", |
936688d7 | 720 | dm_device_name(t->md), type); |
5ae89a87 | 721 | |
899ab445 | 722 | if (ti->limit_swap_bios && !static_key_enabled(&swap_bios_enabled.key)) |
442761fd MS |
723 | static_branch_enable(&swap_bios_enabled); |
724 | ||
1da177e4 LT |
725 | return 0; |
726 | ||
727 | bad: | |
899ab445 MS |
728 | DMERR("%s: %s: %s (%pe)", dm_device_name(t->md), type, ti->error, ERR_PTR(r)); |
729 | dm_put_target_type(ti->type); | |
1da177e4 LT |
730 | return r; |
731 | } | |
732 | ||
498f0103 MS |
733 | /* |
734 | * Target argument parsing helpers. | |
735 | */ | |
86a3238c HM |
736 | static int validate_next_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
737 | unsigned int *value, char **error, unsigned int grouped) | |
498f0103 MS |
738 | { |
739 | const char *arg_str = dm_shift_arg(arg_set); | |
31998ef1 | 740 | char dummy; |
498f0103 MS |
741 | |
742 | if (!arg_str || | |
31998ef1 | 743 | (sscanf(arg_str, "%u%c", value, &dummy) != 1) || |
498f0103 MS |
744 | (*value < arg->min) || |
745 | (*value > arg->max) || | |
746 | (grouped && arg_set->argc < *value)) { | |
747 | *error = arg->error; | |
748 | return -EINVAL; | |
749 | } | |
750 | ||
751 | return 0; | |
752 | } | |
753 | ||
5916a22b | 754 | int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
86a3238c | 755 | unsigned int *value, char **error) |
498f0103 MS |
756 | { |
757 | return validate_next_arg(arg, arg_set, value, error, 0); | |
758 | } | |
759 | EXPORT_SYMBOL(dm_read_arg); | |
760 | ||
5916a22b | 761 | int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set, |
86a3238c | 762 | unsigned int *value, char **error) |
498f0103 MS |
763 | { |
764 | return validate_next_arg(arg, arg_set, value, error, 1); | |
765 | } | |
766 | EXPORT_SYMBOL(dm_read_arg_group); | |
767 | ||
768 | const char *dm_shift_arg(struct dm_arg_set *as) | |
769 | { | |
770 | char *r; | |
771 | ||
772 | if (as->argc) { | |
773 | as->argc--; | |
774 | r = *as->argv; | |
775 | as->argv++; | |
776 | return r; | |
777 | } | |
778 | ||
779 | return NULL; | |
780 | } | |
781 | EXPORT_SYMBOL(dm_shift_arg); | |
782 | ||
86a3238c | 783 | void dm_consume_args(struct dm_arg_set *as, unsigned int num_args) |
498f0103 MS |
784 | { |
785 | BUG_ON(as->argc < num_args); | |
786 | as->argc -= num_args; | |
787 | as->argv += num_args; | |
788 | } | |
789 | EXPORT_SYMBOL(dm_consume_args); | |
790 | ||
7e0d574f | 791 | static bool __table_type_bio_based(enum dm_queue_mode table_type) |
545ed20e TK |
792 | { |
793 | return (table_type == DM_TYPE_BIO_BASED || | |
9c37de29 | 794 | table_type == DM_TYPE_DAX_BIO_BASED); |
545ed20e TK |
795 | } |
796 | ||
7e0d574f | 797 | static bool __table_type_request_based(enum dm_queue_mode table_type) |
15b94a69 | 798 | { |
953923c0 | 799 | return table_type == DM_TYPE_REQUEST_BASED; |
15b94a69 JN |
800 | } |
801 | ||
7e0d574f | 802 | void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type) |
e83068a5 MS |
803 | { |
804 | t->type = type; | |
805 | } | |
806 | EXPORT_SYMBOL_GPL(dm_table_set_type); | |
807 | ||
7bf7eac8 | 808 | /* validate the dax capability of the target device span */ |
7b0800d0 | 809 | static int device_not_dax_capable(struct dm_target *ti, struct dm_dev *dev, |
9c50a98f | 810 | sector_t start, sector_t len, void *data) |
545ed20e | 811 | { |
7b0800d0 CH |
812 | if (dev->dax_dev) |
813 | return false; | |
7bf7eac8 | 814 | |
7b0800d0 CH |
815 | DMDEBUG("%pg: error: dax unsupported by block device", dev->bdev); |
816 | return true; | |
545ed20e TK |
817 | } |
818 | ||
2e9ee095 | 819 | /* Check devices support synchronous DAX */ |
5b0fab50 JX |
820 | static int device_not_dax_synchronous_capable(struct dm_target *ti, struct dm_dev *dev, |
821 | sector_t start, sector_t len, void *data) | |
2e9ee095 | 822 | { |
5b0fab50 | 823 | return !dev->dax_dev || !dax_synchronous(dev->dax_dev); |
2e9ee095 PG |
824 | } |
825 | ||
7b0800d0 | 826 | static bool dm_table_supports_dax(struct dm_table *t, |
564b5c54 | 827 | iterate_devices_callout_fn iterate_fn) |
545ed20e | 828 | { |
545ed20e | 829 | /* Ensure that all targets support DAX. */ |
564b5c54 MS |
830 | for (unsigned int i = 0; i < t->num_targets; i++) { |
831 | struct dm_target *ti = dm_table_get_target(t, i); | |
545ed20e TK |
832 | |
833 | if (!ti->type->direct_access) | |
834 | return false; | |
835 | ||
836 | if (!ti->type->iterate_devices || | |
7b0800d0 | 837 | ti->type->iterate_devices(ti, iterate_fn, NULL)) |
545ed20e TK |
838 | return false; |
839 | } | |
840 | ||
841 | return true; | |
842 | } | |
843 | ||
6ba01df7 MS |
844 | static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev, |
845 | sector_t start, sector_t len, void *data) | |
eaa160ed | 846 | { |
6ba01df7 MS |
847 | struct block_device *bdev = dev->bdev; |
848 | struct request_queue *q = bdev_get_queue(bdev); | |
eaa160ed | 849 | |
6ba01df7 | 850 | /* request-based cannot stack on partitions! */ |
fa01b1e9 | 851 | if (bdev_is_partition(bdev)) |
6ba01df7 | 852 | return false; |
eaa160ed | 853 | |
344e9ffc | 854 | return queue_is_mq(q); |
eaa160ed MS |
855 | } |
856 | ||
e83068a5 | 857 | static int dm_table_determine_type(struct dm_table *t) |
e6ee8c0b | 858 | { |
86a3238c | 859 | unsigned int bio_based = 0, request_based = 0, hybrid = 0; |
564b5c54 | 860 | struct dm_target *ti; |
e83068a5 | 861 | struct list_head *devices = dm_table_get_devices(t); |
7e0d574f | 862 | enum dm_queue_mode live_md_type = dm_get_md_type(t->md); |
e6ee8c0b | 863 | |
e83068a5 MS |
864 | if (t->type != DM_TYPE_NONE) { |
865 | /* target already set the table's type */ | |
c934edad MS |
866 | if (t->type == DM_TYPE_BIO_BASED) { |
867 | /* possibly upgrade to a variant of bio-based */ | |
868 | goto verify_bio_based; | |
22c11858 | 869 | } |
545ed20e | 870 | BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED); |
e83068a5 MS |
871 | goto verify_rq_based; |
872 | } | |
873 | ||
564b5c54 MS |
874 | for (unsigned int i = 0; i < t->num_targets; i++) { |
875 | ti = dm_table_get_target(t, i); | |
876 | if (dm_target_hybrid(ti)) | |
169e2cc2 | 877 | hybrid = 1; |
564b5c54 | 878 | else if (dm_target_request_based(ti)) |
e6ee8c0b KU |
879 | request_based = 1; |
880 | else | |
881 | bio_based = 1; | |
882 | ||
883 | if (bio_based && request_based) { | |
2e84fecf | 884 | DMERR("Inconsistent table: different target types can't be mixed up"); |
e6ee8c0b KU |
885 | return -EINVAL; |
886 | } | |
887 | } | |
888 | ||
169e2cc2 MS |
889 | if (hybrid && !bio_based && !request_based) { |
890 | /* | |
891 | * The targets can work either way. | |
892 | * Determine the type from the live device. | |
893 | * Default to bio-based if device is new. | |
894 | */ | |
15b94a69 | 895 | if (__table_type_request_based(live_md_type)) |
169e2cc2 MS |
896 | request_based = 1; |
897 | else | |
898 | bio_based = 1; | |
899 | } | |
900 | ||
e6ee8c0b | 901 | if (bio_based) { |
c934edad | 902 | verify_bio_based: |
e6ee8c0b KU |
903 | /* We must use this table as bio-based */ |
904 | t->type = DM_TYPE_BIO_BASED; | |
7b0800d0 | 905 | if (dm_table_supports_dax(t, device_not_dax_capable) || |
22c11858 | 906 | (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) { |
545ed20e | 907 | t->type = DM_TYPE_DAX_BIO_BASED; |
22c11858 | 908 | } |
e6ee8c0b KU |
909 | return 0; |
910 | } | |
911 | ||
912 | BUG_ON(!request_based); /* No targets in this table */ | |
913 | ||
e83068a5 MS |
914 | t->type = DM_TYPE_REQUEST_BASED; |
915 | ||
916 | verify_rq_based: | |
65803c20 MS |
917 | /* |
918 | * Request-based dm supports only tables that have a single target now. | |
919 | * To support multiple targets, request splitting support is needed, | |
920 | * and that needs lots of changes in the block-layer. | |
921 | * (e.g. request completion process for partial completion.) | |
922 | */ | |
923 | if (t->num_targets > 1) { | |
9c37de29 | 924 | DMERR("request-based DM doesn't support multiple targets"); |
65803c20 MS |
925 | return -EINVAL; |
926 | } | |
927 | ||
6936c12c MS |
928 | if (list_empty(devices)) { |
929 | int srcu_idx; | |
930 | struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx); | |
931 | ||
6a23e05c JA |
932 | /* inherit live table's type */ |
933 | if (live_table) | |
6936c12c | 934 | t->type = live_table->type; |
6936c12c MS |
935 | dm_put_live_table(t->md, srcu_idx); |
936 | return 0; | |
937 | } | |
938 | ||
564b5c54 MS |
939 | ti = dm_table_get_immutable_target(t); |
940 | if (!ti) { | |
22c11858 MS |
941 | DMERR("table load rejected: immutable target is required"); |
942 | return -EINVAL; | |
564b5c54 | 943 | } else if (ti->max_io_len) { |
22c11858 MS |
944 | DMERR("table load rejected: immutable target that splits IO is not supported"); |
945 | return -EINVAL; | |
946 | } | |
947 | ||
e6ee8c0b | 948 | /* Non-request-stackable devices can't be used for request-based dm */ |
564b5c54 MS |
949 | if (!ti->type->iterate_devices || |
950 | !ti->type->iterate_devices(ti, device_is_rq_stackable, NULL)) { | |
eaa160ed MS |
951 | DMERR("table load rejected: including non-request-stackable devices"); |
952 | return -EINVAL; | |
e5863d9a | 953 | } |
301fc3f5 | 954 | |
e6ee8c0b KU |
955 | return 0; |
956 | } | |
957 | ||
7e0d574f | 958 | enum dm_queue_mode dm_table_get_type(struct dm_table *t) |
e6ee8c0b KU |
959 | { |
960 | return t->type; | |
961 | } | |
962 | ||
36a0456f AK |
963 | struct target_type *dm_table_get_immutable_target_type(struct dm_table *t) |
964 | { | |
965 | return t->immutable_target_type; | |
966 | } | |
967 | ||
16f12266 MS |
968 | struct dm_target *dm_table_get_immutable_target(struct dm_table *t) |
969 | { | |
970 | /* Immutable target is implicitly a singleton */ | |
971 | if (t->num_targets > 1 || | |
972 | !dm_target_is_immutable(t->targets[0].type)) | |
973 | return NULL; | |
974 | ||
975 | return t->targets; | |
976 | } | |
977 | ||
f083b09b MS |
978 | struct dm_target *dm_table_get_wildcard_target(struct dm_table *t) |
979 | { | |
564b5c54 MS |
980 | for (unsigned int i = 0; i < t->num_targets; i++) { |
981 | struct dm_target *ti = dm_table_get_target(t, i); | |
f083b09b | 982 | |
f083b09b MS |
983 | if (dm_target_is_wildcard(ti->type)) |
984 | return ti; | |
985 | } | |
986 | ||
987 | return NULL; | |
988 | } | |
989 | ||
545ed20e TK |
990 | bool dm_table_bio_based(struct dm_table *t) |
991 | { | |
992 | return __table_type_bio_based(dm_table_get_type(t)); | |
993 | } | |
994 | ||
e6ee8c0b KU |
995 | bool dm_table_request_based(struct dm_table *t) |
996 | { | |
15b94a69 | 997 | return __table_type_request_based(dm_table_get_type(t)); |
e5863d9a MS |
998 | } |
999 | ||
9571f829 | 1000 | static bool dm_table_supports_poll(struct dm_table *t); |
cfc97abc | 1001 | |
17e149b8 | 1002 | static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md) |
e6ee8c0b | 1003 | { |
7e0d574f | 1004 | enum dm_queue_mode type = dm_table_get_type(t); |
e810cb78 CH |
1005 | unsigned int per_io_data_size = 0, front_pad, io_front_pad; |
1006 | unsigned int min_pool_size = 0, pool_size; | |
1007 | struct dm_md_mempools *pools; | |
e6ee8c0b | 1008 | |
78d8e58a | 1009 | if (unlikely(type == DM_TYPE_NONE)) { |
43e6c111 | 1010 | DMERR("no table type is set, can't allocate mempools"); |
e6ee8c0b KU |
1011 | return -EINVAL; |
1012 | } | |
1013 | ||
e810cb78 CH |
1014 | pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id); |
1015 | if (!pools) | |
1016 | return -ENOMEM; | |
1017 | ||
1018 | if (type == DM_TYPE_REQUEST_BASED) { | |
1019 | pool_size = dm_get_reserved_rq_based_ios(); | |
1020 | front_pad = offsetof(struct dm_rq_clone_bio_info, clone); | |
1021 | goto init_bs; | |
cfc97abc | 1022 | } |
78d8e58a | 1023 | |
e810cb78 | 1024 | for (unsigned int i = 0; i < t->num_targets; i++) { |
564b5c54 | 1025 | struct dm_target *ti = dm_table_get_target(t, i); |
e6ee8c0b | 1026 | |
e810cb78 CH |
1027 | per_io_data_size = max(per_io_data_size, ti->per_io_data_size); |
1028 | min_pool_size = max(min_pool_size, ti->num_flush_bios); | |
1029 | } | |
1030 | pool_size = max(dm_get_reserved_bio_based_ios(), min_pool_size); | |
1031 | front_pad = roundup(per_io_data_size, | |
1032 | __alignof__(struct dm_target_io)) + DM_TARGET_IO_BIO_OFFSET; | |
1033 | ||
1034 | io_front_pad = roundup(per_io_data_size, | |
1035 | __alignof__(struct dm_io)) + DM_IO_BIO_OFFSET; | |
1036 | if (bioset_init(&pools->io_bs, pool_size, io_front_pad, | |
1037 | dm_table_supports_poll(t) ? BIOSET_PERCPU_CACHE : 0)) | |
1038 | goto out_free_pools; | |
1039 | if (t->integrity_supported && | |
1040 | bioset_integrity_create(&pools->io_bs, pool_size)) | |
1041 | goto out_free_pools; | |
1042 | init_bs: | |
1043 | if (bioset_init(&pools->bs, pool_size, front_pad, 0)) | |
1044 | goto out_free_pools; | |
1045 | if (t->integrity_supported && | |
1046 | bioset_integrity_create(&pools->bs, pool_size)) | |
1047 | goto out_free_pools; | |
1048 | ||
1049 | t->mempools = pools; | |
e6ee8c0b | 1050 | return 0; |
e810cb78 CH |
1051 | |
1052 | out_free_pools: | |
1053 | dm_free_md_mempools(pools); | |
1054 | return -ENOMEM; | |
e6ee8c0b KU |
1055 | } |
1056 | ||
1da177e4 LT |
1057 | static int setup_indexes(struct dm_table *t) |
1058 | { | |
1059 | int i; | |
1060 | unsigned int total = 0; | |
1061 | sector_t *indexes; | |
1062 | ||
1063 | /* allocate the space for *all* the indexes */ | |
1064 | for (i = t->depth - 2; i >= 0; i--) { | |
1065 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
1066 | total += t->counts[i]; | |
1067 | } | |
1068 | ||
7a35693a | 1069 | indexes = kvcalloc(total, NODE_SIZE, GFP_KERNEL); |
1da177e4 LT |
1070 | if (!indexes) |
1071 | return -ENOMEM; | |
1072 | ||
1073 | /* set up internal nodes, bottom-up */ | |
82d601dc | 1074 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
1075 | t->index[i] = indexes; |
1076 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
1077 | setup_btree_index(i, t); | |
1078 | } | |
1079 | ||
1080 | return 0; | |
1081 | } | |
1082 | ||
1083 | /* | |
1084 | * Builds the btree to index the map. | |
1085 | */ | |
26803b9f | 1086 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
1087 | { |
1088 | int r = 0; | |
1089 | unsigned int leaf_nodes; | |
1090 | ||
1da177e4 LT |
1091 | /* how many indexes will the btree have ? */ |
1092 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
1093 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
1094 | ||
1095 | /* leaf layer has already been set up */ | |
1096 | t->counts[t->depth - 1] = leaf_nodes; | |
1097 | t->index[t->depth - 1] = t->highs; | |
1098 | ||
1099 | if (t->depth >= 2) | |
1100 | r = setup_indexes(t); | |
1101 | ||
1102 | return r; | |
1103 | } | |
1104 | ||
25520d55 MP |
1105 | static bool integrity_profile_exists(struct gendisk *disk) |
1106 | { | |
1107 | return !!blk_get_integrity(disk); | |
1108 | } | |
1109 | ||
a63a5cf8 MS |
1110 | /* |
1111 | * Get a disk whose integrity profile reflects the table's profile. | |
a63a5cf8 MS |
1112 | * Returns NULL if integrity support was inconsistent or unavailable. |
1113 | */ | |
43e6c111 | 1114 | static struct gendisk *dm_table_get_integrity_disk(struct dm_table *t) |
a63a5cf8 MS |
1115 | { |
1116 | struct list_head *devices = dm_table_get_devices(t); | |
1117 | struct dm_dev_internal *dd = NULL; | |
1118 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
e2460f2a | 1119 | |
564b5c54 | 1120 | for (unsigned int i = 0; i < t->num_targets; i++) { |
e2460f2a | 1121 | struct dm_target *ti = dm_table_get_target(t, i); |
564b5c54 | 1122 | |
e2460f2a MP |
1123 | if (!dm_target_passes_integrity(ti->type)) |
1124 | goto no_integrity; | |
1125 | } | |
a63a5cf8 MS |
1126 | |
1127 | list_for_each_entry(dd, devices, list) { | |
86f1152b | 1128 | template_disk = dd->dm_dev->bdev->bd_disk; |
25520d55 | 1129 | if (!integrity_profile_exists(template_disk)) |
a63a5cf8 | 1130 | goto no_integrity; |
a63a5cf8 MS |
1131 | else if (prev_disk && |
1132 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
1133 | goto no_integrity; | |
1134 | prev_disk = template_disk; | |
1135 | } | |
1136 | ||
1137 | return template_disk; | |
1138 | ||
1139 | no_integrity: | |
1140 | if (prev_disk) | |
1141 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
1142 | dm_device_name(t->md), | |
1143 | prev_disk->disk_name, | |
1144 | template_disk->disk_name); | |
1145 | return NULL; | |
1146 | } | |
1147 | ||
26803b9f | 1148 | /* |
25520d55 MP |
1149 | * Register the mapped device for blk_integrity support if the |
1150 | * underlying devices have an integrity profile. But all devices may | |
1151 | * not have matching profiles (checking all devices isn't reliable | |
a63a5cf8 | 1152 | * during table load because this table may use other DM device(s) which |
25520d55 MP |
1153 | * must be resumed before they will have an initialized integity |
1154 | * profile). Consequently, stacked DM devices force a 2 stage integrity | |
1155 | * profile validation: First pass during table load, final pass during | |
1156 | * resume. | |
26803b9f | 1157 | */ |
25520d55 | 1158 | static int dm_table_register_integrity(struct dm_table *t) |
26803b9f | 1159 | { |
25520d55 | 1160 | struct mapped_device *md = t->md; |
a63a5cf8 | 1161 | struct gendisk *template_disk = NULL; |
26803b9f | 1162 | |
9b4b5a79 MB |
1163 | /* If target handles integrity itself do not register it here. */ |
1164 | if (t->integrity_added) | |
1165 | return 0; | |
1166 | ||
25520d55 | 1167 | template_disk = dm_table_get_integrity_disk(t); |
a63a5cf8 MS |
1168 | if (!template_disk) |
1169 | return 0; | |
26803b9f | 1170 | |
25520d55 | 1171 | if (!integrity_profile_exists(dm_disk(md))) { |
e83068a5 | 1172 | t->integrity_supported = true; |
25520d55 MP |
1173 | /* |
1174 | * Register integrity profile during table load; we can do | |
1175 | * this because the final profile must match during resume. | |
1176 | */ | |
1177 | blk_integrity_register(dm_disk(md), | |
1178 | blk_get_integrity(template_disk)); | |
1179 | return 0; | |
a63a5cf8 MS |
1180 | } |
1181 | ||
1182 | /* | |
25520d55 | 1183 | * If DM device already has an initialized integrity |
a63a5cf8 MS |
1184 | * profile the new profile should not conflict. |
1185 | */ | |
25520d55 | 1186 | if (blk_integrity_compare(dm_disk(md), template_disk) < 0) { |
2e84fecf | 1187 | DMERR("%s: conflict with existing integrity profile: %s profile mismatch", |
43e6c111 MP |
1188 | dm_device_name(t->md), |
1189 | template_disk->disk_name); | |
a63a5cf8 MS |
1190 | return 1; |
1191 | } | |
1192 | ||
25520d55 | 1193 | /* Preserve existing integrity profile */ |
e83068a5 | 1194 | t->integrity_supported = true; |
26803b9f WD |
1195 | return 0; |
1196 | } | |
1197 | ||
aa6ce87a ST |
1198 | #ifdef CONFIG_BLK_INLINE_ENCRYPTION |
1199 | ||
cb77cb5a EB |
1200 | struct dm_crypto_profile { |
1201 | struct blk_crypto_profile profile; | |
aa6ce87a ST |
1202 | struct mapped_device *md; |
1203 | }; | |
1204 | ||
9355a9eb ST |
1205 | static int dm_keyslot_evict_callback(struct dm_target *ti, struct dm_dev *dev, |
1206 | sector_t start, sector_t len, void *data) | |
1207 | { | |
70493a63 | 1208 | const struct blk_crypto_key *key = data; |
9355a9eb | 1209 | |
70493a63 | 1210 | blk_crypto_evict_key(dev->bdev, key); |
9355a9eb ST |
1211 | return 0; |
1212 | } | |
1213 | ||
1214 | /* | |
1215 | * When an inline encryption key is evicted from a device-mapper device, evict | |
1216 | * it from all the underlying devices. | |
1217 | */ | |
cb77cb5a | 1218 | static int dm_keyslot_evict(struct blk_crypto_profile *profile, |
9355a9eb ST |
1219 | const struct blk_crypto_key *key, unsigned int slot) |
1220 | { | |
cb77cb5a EB |
1221 | struct mapped_device *md = |
1222 | container_of(profile, struct dm_crypto_profile, profile)->md; | |
9355a9eb ST |
1223 | struct dm_table *t; |
1224 | int srcu_idx; | |
9355a9eb ST |
1225 | |
1226 | t = dm_get_live_table(md, &srcu_idx); | |
1227 | if (!t) | |
1228 | return 0; | |
564b5c54 MS |
1229 | |
1230 | for (unsigned int i = 0; i < t->num_targets; i++) { | |
1231 | struct dm_target *ti = dm_table_get_target(t, i); | |
1232 | ||
9355a9eb ST |
1233 | if (!ti->type->iterate_devices) |
1234 | continue; | |
70493a63 EB |
1235 | ti->type->iterate_devices(ti, dm_keyslot_evict_callback, |
1236 | (void *)key); | |
9355a9eb | 1237 | } |
564b5c54 | 1238 | |
9355a9eb | 1239 | dm_put_live_table(md, srcu_idx); |
70493a63 | 1240 | return 0; |
9355a9eb ST |
1241 | } |
1242 | ||
cb77cb5a EB |
1243 | static int |
1244 | device_intersect_crypto_capabilities(struct dm_target *ti, struct dm_dev *dev, | |
1245 | sector_t start, sector_t len, void *data) | |
aa6ce87a | 1246 | { |
cb77cb5a EB |
1247 | struct blk_crypto_profile *parent = data; |
1248 | struct blk_crypto_profile *child = | |
1249 | bdev_get_queue(dev->bdev)->crypto_profile; | |
aa6ce87a | 1250 | |
cb77cb5a | 1251 | blk_crypto_intersect_capabilities(parent, child); |
aa6ce87a ST |
1252 | return 0; |
1253 | } | |
1254 | ||
cb77cb5a | 1255 | void dm_destroy_crypto_profile(struct blk_crypto_profile *profile) |
aa6ce87a | 1256 | { |
cb77cb5a EB |
1257 | struct dm_crypto_profile *dmcp = container_of(profile, |
1258 | struct dm_crypto_profile, | |
1259 | profile); | |
aa6ce87a | 1260 | |
cb77cb5a | 1261 | if (!profile) |
aa6ce87a ST |
1262 | return; |
1263 | ||
cb77cb5a EB |
1264 | blk_crypto_profile_destroy(profile); |
1265 | kfree(dmcp); | |
aa6ce87a ST |
1266 | } |
1267 | ||
cb77cb5a | 1268 | static void dm_table_destroy_crypto_profile(struct dm_table *t) |
aa6ce87a | 1269 | { |
cb77cb5a EB |
1270 | dm_destroy_crypto_profile(t->crypto_profile); |
1271 | t->crypto_profile = NULL; | |
aa6ce87a ST |
1272 | } |
1273 | ||
1274 | /* | |
cb77cb5a EB |
1275 | * Constructs and initializes t->crypto_profile with a crypto profile that |
1276 | * represents the common set of crypto capabilities of the devices described by | |
1277 | * the dm_table. However, if the constructed crypto profile doesn't support all | |
1278 | * crypto capabilities that are supported by the current mapped_device, it | |
1279 | * returns an error instead, since we don't support removing crypto capabilities | |
1280 | * on table changes. Finally, if the constructed crypto profile is "empty" (has | |
1281 | * no crypto capabilities at all), it just sets t->crypto_profile to NULL. | |
aa6ce87a | 1282 | */ |
cb77cb5a | 1283 | static int dm_table_construct_crypto_profile(struct dm_table *t) |
aa6ce87a | 1284 | { |
cb77cb5a EB |
1285 | struct dm_crypto_profile *dmcp; |
1286 | struct blk_crypto_profile *profile; | |
aa6ce87a | 1287 | unsigned int i; |
cb77cb5a | 1288 | bool empty_profile = true; |
aa6ce87a | 1289 | |
cb77cb5a EB |
1290 | dmcp = kmalloc(sizeof(*dmcp), GFP_KERNEL); |
1291 | if (!dmcp) | |
aa6ce87a | 1292 | return -ENOMEM; |
cb77cb5a | 1293 | dmcp->md = t->md; |
aa6ce87a | 1294 | |
cb77cb5a EB |
1295 | profile = &dmcp->profile; |
1296 | blk_crypto_profile_init(profile, 0); | |
1297 | profile->ll_ops.keyslot_evict = dm_keyslot_evict; | |
1298 | profile->max_dun_bytes_supported = UINT_MAX; | |
1299 | memset(profile->modes_supported, 0xFF, | |
1300 | sizeof(profile->modes_supported)); | |
aa6ce87a | 1301 | |
2aec377a | 1302 | for (i = 0; i < t->num_targets; i++) { |
564b5c54 | 1303 | struct dm_target *ti = dm_table_get_target(t, i); |
aa6ce87a ST |
1304 | |
1305 | if (!dm_target_passes_crypto(ti->type)) { | |
cb77cb5a | 1306 | blk_crypto_intersect_capabilities(profile, NULL); |
aa6ce87a ST |
1307 | break; |
1308 | } | |
1309 | if (!ti->type->iterate_devices) | |
1310 | continue; | |
cb77cb5a EB |
1311 | ti->type->iterate_devices(ti, |
1312 | device_intersect_crypto_capabilities, | |
1313 | profile); | |
aa6ce87a ST |
1314 | } |
1315 | ||
cb77cb5a EB |
1316 | if (t->md->queue && |
1317 | !blk_crypto_has_capabilities(profile, | |
1318 | t->md->queue->crypto_profile)) { | |
43e6c111 | 1319 | DMERR("Inline encryption capabilities of new DM table were more restrictive than the old table's. This is not supported!"); |
cb77cb5a | 1320 | dm_destroy_crypto_profile(profile); |
aa6ce87a ST |
1321 | return -EINVAL; |
1322 | } | |
1323 | ||
1324 | /* | |
cb77cb5a EB |
1325 | * If the new profile doesn't actually support any crypto capabilities, |
1326 | * we may as well represent it with a NULL profile. | |
aa6ce87a | 1327 | */ |
cb77cb5a EB |
1328 | for (i = 0; i < ARRAY_SIZE(profile->modes_supported); i++) { |
1329 | if (profile->modes_supported[i]) { | |
1330 | empty_profile = false; | |
aa6ce87a ST |
1331 | break; |
1332 | } | |
1333 | } | |
1334 | ||
cb77cb5a EB |
1335 | if (empty_profile) { |
1336 | dm_destroy_crypto_profile(profile); | |
1337 | profile = NULL; | |
aa6ce87a ST |
1338 | } |
1339 | ||
1340 | /* | |
cb77cb5a EB |
1341 | * t->crypto_profile is only set temporarily while the table is being |
1342 | * set up, and it gets set to NULL after the profile has been | |
1343 | * transferred to the request_queue. | |
aa6ce87a | 1344 | */ |
cb77cb5a | 1345 | t->crypto_profile = profile; |
aa6ce87a ST |
1346 | |
1347 | return 0; | |
1348 | } | |
1349 | ||
cb77cb5a EB |
1350 | static void dm_update_crypto_profile(struct request_queue *q, |
1351 | struct dm_table *t) | |
aa6ce87a | 1352 | { |
cb77cb5a | 1353 | if (!t->crypto_profile) |
aa6ce87a ST |
1354 | return; |
1355 | ||
cb77cb5a EB |
1356 | /* Make the crypto profile less restrictive. */ |
1357 | if (!q->crypto_profile) { | |
1358 | blk_crypto_register(t->crypto_profile, q); | |
aa6ce87a | 1359 | } else { |
cb77cb5a EB |
1360 | blk_crypto_update_capabilities(q->crypto_profile, |
1361 | t->crypto_profile); | |
1362 | dm_destroy_crypto_profile(t->crypto_profile); | |
aa6ce87a | 1363 | } |
cb77cb5a | 1364 | t->crypto_profile = NULL; |
aa6ce87a ST |
1365 | } |
1366 | ||
1367 | #else /* CONFIG_BLK_INLINE_ENCRYPTION */ | |
1368 | ||
cb77cb5a | 1369 | static int dm_table_construct_crypto_profile(struct dm_table *t) |
aa6ce87a ST |
1370 | { |
1371 | return 0; | |
1372 | } | |
1373 | ||
cb77cb5a | 1374 | void dm_destroy_crypto_profile(struct blk_crypto_profile *profile) |
aa6ce87a ST |
1375 | { |
1376 | } | |
1377 | ||
cb77cb5a | 1378 | static void dm_table_destroy_crypto_profile(struct dm_table *t) |
aa6ce87a ST |
1379 | { |
1380 | } | |
1381 | ||
cb77cb5a EB |
1382 | static void dm_update_crypto_profile(struct request_queue *q, |
1383 | struct dm_table *t) | |
aa6ce87a ST |
1384 | { |
1385 | } | |
1386 | ||
1387 | #endif /* !CONFIG_BLK_INLINE_ENCRYPTION */ | |
1388 | ||
26803b9f WD |
1389 | /* |
1390 | * Prepares the table for use by building the indices, | |
1391 | * setting the type, and allocating mempools. | |
1392 | */ | |
1393 | int dm_table_complete(struct dm_table *t) | |
1394 | { | |
1395 | int r; | |
1396 | ||
e83068a5 | 1397 | r = dm_table_determine_type(t); |
26803b9f | 1398 | if (r) { |
e83068a5 | 1399 | DMERR("unable to determine table type"); |
26803b9f WD |
1400 | return r; |
1401 | } | |
1402 | ||
1403 | r = dm_table_build_index(t); | |
1404 | if (r) { | |
1405 | DMERR("unable to build btrees"); | |
1406 | return r; | |
1407 | } | |
1408 | ||
25520d55 | 1409 | r = dm_table_register_integrity(t); |
26803b9f WD |
1410 | if (r) { |
1411 | DMERR("could not register integrity profile."); | |
1412 | return r; | |
1413 | } | |
1414 | ||
cb77cb5a | 1415 | r = dm_table_construct_crypto_profile(t); |
aa6ce87a | 1416 | if (r) { |
cb77cb5a | 1417 | DMERR("could not construct crypto profile."); |
aa6ce87a ST |
1418 | return r; |
1419 | } | |
1420 | ||
17e149b8 | 1421 | r = dm_table_alloc_md_mempools(t, t->md); |
26803b9f WD |
1422 | if (r) |
1423 | DMERR("unable to allocate mempools"); | |
1424 | ||
1425 | return r; | |
1426 | } | |
1427 | ||
48c9c27b | 1428 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1429 | void dm_table_event_callback(struct dm_table *t, |
1430 | void (*fn)(void *), void *context) | |
1431 | { | |
48c9c27b | 1432 | mutex_lock(&_event_lock); |
1da177e4 LT |
1433 | t->event_fn = fn; |
1434 | t->event_context = context; | |
48c9c27b | 1435 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1436 | } |
1437 | ||
1438 | void dm_table_event(struct dm_table *t) | |
1439 | { | |
48c9c27b | 1440 | mutex_lock(&_event_lock); |
1da177e4 LT |
1441 | if (t->event_fn) |
1442 | t->event_fn(t->event_context); | |
48c9c27b | 1443 | mutex_unlock(&_event_lock); |
1da177e4 | 1444 | } |
08649012 | 1445 | EXPORT_SYMBOL(dm_table_event); |
1da177e4 | 1446 | |
1cfd5d33 | 1447 | inline sector_t dm_table_get_size(struct dm_table *t) |
1da177e4 LT |
1448 | { |
1449 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1450 | } | |
08649012 | 1451 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1452 | |
1da177e4 LT |
1453 | /* |
1454 | * Search the btree for the correct target. | |
512875bd | 1455 | * |
123d87d5 | 1456 | * Caller should check returned pointer for NULL |
512875bd | 1457 | * to trap I/O beyond end of device. |
1da177e4 LT |
1458 | */ |
1459 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1460 | { | |
1461 | unsigned int l, n = 0, k = 0; | |
1462 | sector_t *node; | |
1463 | ||
1cfd5d33 | 1464 | if (unlikely(sector >= dm_table_get_size(t))) |
123d87d5 | 1465 | return NULL; |
1cfd5d33 | 1466 | |
1da177e4 LT |
1467 | for (l = 0; l < t->depth; l++) { |
1468 | n = get_child(n, k); | |
1469 | node = get_node(t, l, n); | |
1470 | ||
1471 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1472 | if (node[k] >= sector) | |
1473 | break; | |
1474 | } | |
1475 | ||
1476 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1477 | } | |
1478 | ||
b99fdcdc ML |
1479 | static int device_not_poll_capable(struct dm_target *ti, struct dm_dev *dev, |
1480 | sector_t start, sector_t len, void *data) | |
1481 | { | |
1482 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1483 | ||
1484 | return !test_bit(QUEUE_FLAG_POLL, &q->queue_flags); | |
1485 | } | |
1486 | ||
a4c8dd9c JX |
1487 | /* |
1488 | * type->iterate_devices() should be called when the sanity check needs to | |
1489 | * iterate and check all underlying data devices. iterate_devices() will | |
1490 | * iterate all underlying data devices until it encounters a non-zero return | |
1491 | * code, returned by whether the input iterate_devices_callout_fn, or | |
1492 | * iterate_devices() itself internally. | |
1493 | * | |
1494 | * For some target type (e.g. dm-stripe), one call of iterate_devices() may | |
1495 | * iterate multiple underlying devices internally, in which case a non-zero | |
1496 | * return code returned by iterate_devices_callout_fn will stop the iteration | |
1497 | * in advance. | |
1498 | * | |
1499 | * Cases requiring _any_ underlying device supporting some kind of attribute, | |
1500 | * should use the iteration structure like dm_table_any_dev_attr(), or call | |
1501 | * it directly. @func should handle semantics of positive examples, e.g. | |
1502 | * capable of something. | |
1503 | * | |
1504 | * Cases requiring _all_ underlying devices supporting some kind of attribute, | |
1505 | * should use the iteration structure like dm_table_supports_nowait() or | |
1506 | * dm_table_supports_discards(). Or introduce dm_table_all_devs_attr() that | |
1507 | * uses an @anti_func that handle semantics of counter examples, e.g. not | |
24f6b603 | 1508 | * capable of something. So: return !dm_table_any_dev_attr(t, anti_func, data); |
a4c8dd9c JX |
1509 | */ |
1510 | static bool dm_table_any_dev_attr(struct dm_table *t, | |
24f6b603 | 1511 | iterate_devices_callout_fn func, void *data) |
a4c8dd9c | 1512 | { |
564b5c54 MS |
1513 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1514 | struct dm_target *ti = dm_table_get_target(t, i); | |
a4c8dd9c JX |
1515 | |
1516 | if (ti->type->iterate_devices && | |
24f6b603 | 1517 | ti->type->iterate_devices(ti, func, data)) |
a4c8dd9c | 1518 | return true; |
255e2646 | 1519 | } |
a4c8dd9c JX |
1520 | |
1521 | return false; | |
1522 | } | |
1523 | ||
3ae70656 MS |
1524 | static int count_device(struct dm_target *ti, struct dm_dev *dev, |
1525 | sector_t start, sector_t len, void *data) | |
1526 | { | |
86a3238c | 1527 | unsigned int *num_devices = data; |
3ae70656 MS |
1528 | |
1529 | (*num_devices)++; | |
1530 | ||
1531 | return 0; | |
1532 | } | |
1533 | ||
9571f829 | 1534 | static bool dm_table_supports_poll(struct dm_table *t) |
b99fdcdc | 1535 | { |
564b5c54 MS |
1536 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1537 | struct dm_target *ti = dm_table_get_target(t, i); | |
9571f829 MS |
1538 | |
1539 | if (!ti->type->iterate_devices || | |
1540 | ti->type->iterate_devices(ti, device_not_poll_capable, NULL)) | |
1541 | return false; | |
1542 | } | |
1543 | ||
1544 | return true; | |
b99fdcdc ML |
1545 | } |
1546 | ||
3ae70656 MS |
1547 | /* |
1548 | * Check whether a table has no data devices attached using each | |
1549 | * target's iterate_devices method. | |
1550 | * Returns false if the result is unknown because a target doesn't | |
1551 | * support iterate_devices. | |
1552 | */ | |
564b5c54 | 1553 | bool dm_table_has_no_data_devices(struct dm_table *t) |
3ae70656 | 1554 | { |
564b5c54 MS |
1555 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1556 | struct dm_target *ti = dm_table_get_target(t, i); | |
86a3238c | 1557 | unsigned int num_devices = 0; |
3ae70656 MS |
1558 | |
1559 | if (!ti->type->iterate_devices) | |
1560 | return false; | |
1561 | ||
1562 | ti->type->iterate_devices(ti, count_device, &num_devices); | |
1563 | if (num_devices) | |
1564 | return false; | |
1565 | } | |
1566 | ||
1567 | return true; | |
1568 | } | |
1569 | ||
24f6b603 JX |
1570 | static int device_not_zoned_model(struct dm_target *ti, struct dm_dev *dev, |
1571 | sector_t start, sector_t len, void *data) | |
dd88d313 DLM |
1572 | { |
1573 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1574 | enum blk_zoned_model *zoned_model = data; | |
1575 | ||
cccb493c | 1576 | return blk_queue_zoned_model(q) != *zoned_model; |
dd88d313 DLM |
1577 | } |
1578 | ||
2d669ceb SK |
1579 | /* |
1580 | * Check the device zoned model based on the target feature flag. If the target | |
1581 | * has the DM_TARGET_ZONED_HM feature flag set, host-managed zoned devices are | |
1582 | * also accepted but all devices must have the same zoned model. If the target | |
1583 | * has the DM_TARGET_MIXED_ZONED_MODEL feature set, the devices can have any | |
1584 | * zoned model with all zoned devices having the same zone size. | |
1585 | */ | |
dd88d313 DLM |
1586 | static bool dm_table_supports_zoned_model(struct dm_table *t, |
1587 | enum blk_zoned_model zoned_model) | |
1588 | { | |
564b5c54 MS |
1589 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1590 | struct dm_target *ti = dm_table_get_target(t, i); | |
dd88d313 | 1591 | |
2d669ceb SK |
1592 | if (dm_target_supports_zoned_hm(ti->type)) { |
1593 | if (!ti->type->iterate_devices || | |
1594 | ti->type->iterate_devices(ti, device_not_zoned_model, | |
1595 | &zoned_model)) | |
1596 | return false; | |
1597 | } else if (!dm_target_supports_mixed_zoned_model(ti->type)) { | |
1598 | if (zoned_model == BLK_ZONED_HM) | |
1599 | return false; | |
1600 | } | |
dd88d313 DLM |
1601 | } |
1602 | ||
1603 | return true; | |
1604 | } | |
1605 | ||
24f6b603 JX |
1606 | static int device_not_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev, |
1607 | sector_t start, sector_t len, void *data) | |
dd88d313 | 1608 | { |
dd88d313 DLM |
1609 | unsigned int *zone_sectors = data; |
1610 | ||
edd1dbc8 | 1611 | if (!bdev_is_zoned(dev->bdev)) |
2d669ceb | 1612 | return 0; |
de71973c | 1613 | return bdev_zone_sectors(dev->bdev) != *zone_sectors; |
dd88d313 DLM |
1614 | } |
1615 | ||
2d669ceb SK |
1616 | /* |
1617 | * Check consistency of zoned model and zone sectors across all targets. For | |
1618 | * zone sectors, if the destination device is a zoned block device, it shall | |
1619 | * have the specified zone_sectors. | |
1620 | */ | |
564b5c54 | 1621 | static int validate_hardware_zoned_model(struct dm_table *t, |
dd88d313 DLM |
1622 | enum blk_zoned_model zoned_model, |
1623 | unsigned int zone_sectors) | |
1624 | { | |
1625 | if (zoned_model == BLK_ZONED_NONE) | |
1626 | return 0; | |
1627 | ||
564b5c54 | 1628 | if (!dm_table_supports_zoned_model(t, zoned_model)) { |
dd88d313 | 1629 | DMERR("%s: zoned model is not consistent across all devices", |
564b5c54 | 1630 | dm_device_name(t->md)); |
dd88d313 DLM |
1631 | return -EINVAL; |
1632 | } | |
1633 | ||
1634 | /* Check zone size validity and compatibility */ | |
1635 | if (!zone_sectors || !is_power_of_2(zone_sectors)) | |
1636 | return -EINVAL; | |
1637 | ||
564b5c54 | 1638 | if (dm_table_any_dev_attr(t, device_not_matches_zone_sectors, &zone_sectors)) { |
2d669ceb | 1639 | DMERR("%s: zone sectors is not consistent across all zoned devices", |
564b5c54 | 1640 | dm_device_name(t->md)); |
dd88d313 DLM |
1641 | return -EINVAL; |
1642 | } | |
1643 | ||
1644 | return 0; | |
1645 | } | |
1646 | ||
754c5fc7 MS |
1647 | /* |
1648 | * Establish the new table's queue_limits and validate them. | |
1649 | */ | |
564b5c54 | 1650 | int dm_calculate_queue_limits(struct dm_table *t, |
754c5fc7 MS |
1651 | struct queue_limits *limits) |
1652 | { | |
754c5fc7 | 1653 | struct queue_limits ti_limits; |
dd88d313 DLM |
1654 | enum blk_zoned_model zoned_model = BLK_ZONED_NONE; |
1655 | unsigned int zone_sectors = 0; | |
754c5fc7 | 1656 | |
b1bd055d | 1657 | blk_set_stacking_limits(limits); |
754c5fc7 | 1658 | |
564b5c54 MS |
1659 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1660 | struct dm_target *ti = dm_table_get_target(t, i); | |
754c5fc7 | 1661 | |
564b5c54 | 1662 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 | 1663 | |
85c938e8 MP |
1664 | if (!ti->type->iterate_devices) { |
1665 | /* Set I/O hints portion of queue limits */ | |
1666 | if (ti->type->io_hints) | |
1667 | ti->type->io_hints(ti, &ti_limits); | |
754c5fc7 | 1668 | goto combine_limits; |
85c938e8 | 1669 | } |
754c5fc7 MS |
1670 | |
1671 | /* | |
1672 | * Combine queue limits of all the devices this target uses. | |
1673 | */ | |
1674 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1675 | &ti_limits); | |
1676 | ||
dd88d313 DLM |
1677 | if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) { |
1678 | /* | |
1679 | * After stacking all limits, validate all devices | |
1680 | * in table support this zoned model and zone sectors. | |
1681 | */ | |
1682 | zoned_model = ti_limits.zoned; | |
1683 | zone_sectors = ti_limits.chunk_sectors; | |
1684 | } | |
1685 | ||
40bea431 MS |
1686 | /* Set I/O hints portion of queue limits */ |
1687 | if (ti->type->io_hints) | |
1688 | ti->type->io_hints(ti, &ti_limits); | |
1689 | ||
754c5fc7 MS |
1690 | /* |
1691 | * Check each device area is consistent with the target's | |
1692 | * overall queue limits. | |
1693 | */ | |
f6a1ed10 MP |
1694 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1695 | &ti_limits)) | |
754c5fc7 MS |
1696 | return -EINVAL; |
1697 | ||
1698 | combine_limits: | |
1699 | /* | |
1700 | * Merge this target's queue limits into the overall limits | |
1701 | * for the table. | |
1702 | */ | |
1703 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
2e84fecf | 1704 | DMWARN("%s: adding target device (start sect %llu len %llu) " |
b27d7f16 | 1705 | "caused an alignment inconsistency", |
564b5c54 | 1706 | dm_device_name(t->md), |
754c5fc7 MS |
1707 | (unsigned long long) ti->begin, |
1708 | (unsigned long long) ti->len); | |
1709 | } | |
1710 | ||
dd88d313 DLM |
1711 | /* |
1712 | * Verify that the zoned model and zone sectors, as determined before | |
1713 | * any .io_hints override, are the same across all devices in the table. | |
1714 | * - this is especially relevant if .io_hints is emulating a disk-managed | |
1715 | * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices. | |
1716 | * BUT... | |
1717 | */ | |
1718 | if (limits->zoned != BLK_ZONED_NONE) { | |
1719 | /* | |
1720 | * ...IF the above limits stacking determined a zoned model | |
1721 | * validate that all of the table's devices conform to it. | |
1722 | */ | |
1723 | zoned_model = limits->zoned; | |
1724 | zone_sectors = limits->chunk_sectors; | |
1725 | } | |
564b5c54 | 1726 | if (validate_hardware_zoned_model(t, zoned_model, zone_sectors)) |
dd88d313 DLM |
1727 | return -EINVAL; |
1728 | ||
564b5c54 | 1729 | return validate_hardware_logical_block_alignment(t, limits); |
754c5fc7 MS |
1730 | } |
1731 | ||
9c47008d | 1732 | /* |
25520d55 MP |
1733 | * Verify that all devices have an integrity profile that matches the |
1734 | * DM device's registered integrity profile. If the profiles don't | |
1735 | * match then unregister the DM device's integrity profile. | |
9c47008d | 1736 | */ |
25520d55 | 1737 | static void dm_table_verify_integrity(struct dm_table *t) |
9c47008d | 1738 | { |
a63a5cf8 | 1739 | struct gendisk *template_disk = NULL; |
9c47008d | 1740 | |
9b4b5a79 MB |
1741 | if (t->integrity_added) |
1742 | return; | |
1743 | ||
25520d55 MP |
1744 | if (t->integrity_supported) { |
1745 | /* | |
1746 | * Verify that the original integrity profile | |
1747 | * matches all the devices in this table. | |
1748 | */ | |
1749 | template_disk = dm_table_get_integrity_disk(t); | |
1750 | if (template_disk && | |
1751 | blk_integrity_compare(dm_disk(t->md), template_disk) >= 0) | |
1752 | return; | |
1753 | } | |
9c47008d | 1754 | |
25520d55 | 1755 | if (integrity_profile_exists(dm_disk(t->md))) { |
876fbba1 MS |
1756 | DMWARN("%s: unable to establish an integrity profile", |
1757 | dm_device_name(t->md)); | |
25520d55 MP |
1758 | blk_integrity_unregister(dm_disk(t->md)); |
1759 | } | |
9c47008d MP |
1760 | } |
1761 | ||
ed8b752b MS |
1762 | static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, |
1763 | sector_t start, sector_t len, void *data) | |
1764 | { | |
c888a8f9 | 1765 | unsigned long flush = (unsigned long) data; |
ed8b752b MS |
1766 | struct request_queue *q = bdev_get_queue(dev->bdev); |
1767 | ||
cccb493c | 1768 | return (q->queue_flags & flush); |
ed8b752b MS |
1769 | } |
1770 | ||
c888a8f9 | 1771 | static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush) |
ed8b752b | 1772 | { |
ed8b752b MS |
1773 | /* |
1774 | * Require at least one underlying device to support flushes. | |
1775 | * t->devices includes internal dm devices such as mirror logs | |
1776 | * so we need to use iterate_devices here, which targets | |
1777 | * supporting flushes must provide. | |
1778 | */ | |
564b5c54 MS |
1779 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1780 | struct dm_target *ti = dm_table_get_target(t, i); | |
ed8b752b | 1781 | |
55a62eef | 1782 | if (!ti->num_flush_bios) |
ed8b752b MS |
1783 | continue; |
1784 | ||
0e9c24ed | 1785 | if (ti->flush_supported) |
7f61f5a0 | 1786 | return true; |
0e9c24ed | 1787 | |
ed8b752b | 1788 | if (ti->type->iterate_devices && |
c888a8f9 | 1789 | ti->type->iterate_devices(ti, device_flush_capable, (void *) flush)) |
7f61f5a0 | 1790 | return true; |
ed8b752b MS |
1791 | } |
1792 | ||
7f61f5a0 | 1793 | return false; |
ed8b752b MS |
1794 | } |
1795 | ||
273752c9 VG |
1796 | static int device_dax_write_cache_enabled(struct dm_target *ti, |
1797 | struct dm_dev *dev, sector_t start, | |
1798 | sector_t len, void *data) | |
1799 | { | |
1800 | struct dax_device *dax_dev = dev->dax_dev; | |
1801 | ||
1802 | if (!dax_dev) | |
1803 | return false; | |
1804 | ||
1805 | if (dax_write_cache_enabled(dax_dev)) | |
1806 | return true; | |
1807 | return false; | |
1808 | } | |
1809 | ||
a4c8dd9c JX |
1810 | static int device_is_rotational(struct dm_target *ti, struct dm_dev *dev, |
1811 | sector_t start, sector_t len, void *data) | |
4693c966 | 1812 | { |
10f0d2a5 | 1813 | return !bdev_nonrot(dev->bdev); |
4693c966 MSB |
1814 | } |
1815 | ||
c3c4555e MB |
1816 | static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, |
1817 | sector_t start, sector_t len, void *data) | |
1818 | { | |
1819 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1820 | ||
cccb493c | 1821 | return !blk_queue_add_random(q); |
c3c4555e MB |
1822 | } |
1823 | ||
ac62d620 CH |
1824 | static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev, |
1825 | sector_t start, sector_t len, void *data) | |
1826 | { | |
1827 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1828 | ||
cccb493c | 1829 | return !q->limits.max_write_zeroes_sectors; |
ac62d620 CH |
1830 | } |
1831 | ||
1832 | static bool dm_table_supports_write_zeroes(struct dm_table *t) | |
1833 | { | |
564b5c54 MS |
1834 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1835 | struct dm_target *ti = dm_table_get_target(t, i); | |
ac62d620 CH |
1836 | |
1837 | if (!ti->num_write_zeroes_bios) | |
1838 | return false; | |
1839 | ||
1840 | if (!ti->type->iterate_devices || | |
1841 | ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL)) | |
1842 | return false; | |
1843 | } | |
1844 | ||
1845 | return true; | |
1846 | } | |
1847 | ||
6abc4946 KK |
1848 | static int device_not_nowait_capable(struct dm_target *ti, struct dm_dev *dev, |
1849 | sector_t start, sector_t len, void *data) | |
1850 | { | |
568ec936 | 1851 | return !bdev_nowait(dev->bdev); |
6abc4946 KK |
1852 | } |
1853 | ||
1854 | static bool dm_table_supports_nowait(struct dm_table *t) | |
1855 | { | |
564b5c54 MS |
1856 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1857 | struct dm_target *ti = dm_table_get_target(t, i); | |
6abc4946 KK |
1858 | |
1859 | if (!dm_target_supports_nowait(ti->type)) | |
1860 | return false; | |
1861 | ||
1862 | if (!ti->type->iterate_devices || | |
1863 | ti->type->iterate_devices(ti, device_not_nowait_capable, NULL)) | |
1864 | return false; | |
1865 | } | |
1866 | ||
1867 | return true; | |
1868 | } | |
1869 | ||
8a74d29d MS |
1870 | static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1871 | sector_t start, sector_t len, void *data) | |
a7ffb6a5 | 1872 | { |
70200574 | 1873 | return !bdev_max_discard_sectors(dev->bdev); |
a7ffb6a5 MP |
1874 | } |
1875 | ||
1876 | static bool dm_table_supports_discards(struct dm_table *t) | |
1877 | { | |
564b5c54 MS |
1878 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1879 | struct dm_target *ti = dm_table_get_target(t, i); | |
a7ffb6a5 MP |
1880 | |
1881 | if (!ti->num_discard_bios) | |
8a74d29d | 1882 | return false; |
a7ffb6a5 | 1883 | |
8a74d29d MS |
1884 | /* |
1885 | * Either the target provides discard support (as implied by setting | |
1886 | * 'discards_supported') or it relies on _all_ data devices having | |
1887 | * discard support. | |
1888 | */ | |
1889 | if (!ti->discards_supported && | |
1890 | (!ti->type->iterate_devices || | |
1891 | ti->type->iterate_devices(ti, device_not_discard_capable, NULL))) | |
1892 | return false; | |
a7ffb6a5 MP |
1893 | } |
1894 | ||
8a74d29d | 1895 | return true; |
a7ffb6a5 MP |
1896 | } |
1897 | ||
00716545 DS |
1898 | static int device_not_secure_erase_capable(struct dm_target *ti, |
1899 | struct dm_dev *dev, sector_t start, | |
1900 | sector_t len, void *data) | |
1901 | { | |
44abff2c | 1902 | return !bdev_max_secure_erase_sectors(dev->bdev); |
00716545 DS |
1903 | } |
1904 | ||
1905 | static bool dm_table_supports_secure_erase(struct dm_table *t) | |
1906 | { | |
564b5c54 MS |
1907 | for (unsigned int i = 0; i < t->num_targets; i++) { |
1908 | struct dm_target *ti = dm_table_get_target(t, i); | |
00716545 DS |
1909 | |
1910 | if (!ti->num_secure_erase_bios) | |
1911 | return false; | |
1912 | ||
1913 | if (!ti->type->iterate_devices || | |
1914 | ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL)) | |
1915 | return false; | |
1916 | } | |
1917 | ||
1918 | return true; | |
1919 | } | |
1920 | ||
eb40c0ac ID |
1921 | static int device_requires_stable_pages(struct dm_target *ti, |
1922 | struct dm_dev *dev, sector_t start, | |
1923 | sector_t len, void *data) | |
1924 | { | |
36d25489 | 1925 | return bdev_stable_writes(dev->bdev); |
eb40c0ac ID |
1926 | } |
1927 | ||
bb37d772 DLM |
1928 | int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1929 | struct queue_limits *limits) | |
1da177e4 | 1930 | { |
519a7e16 | 1931 | bool wc = false, fua = false; |
bb37d772 | 1932 | int r; |
ed8b752b | 1933 | |
1da177e4 | 1934 | /* |
1197764e | 1935 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1936 | */ |
754c5fc7 | 1937 | q->limits = *limits; |
c9a3f6d6 | 1938 | |
6abc4946 KK |
1939 | if (dm_table_supports_nowait(t)) |
1940 | blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q); | |
1941 | else | |
1942 | blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, q); | |
1943 | ||
5d47c89f | 1944 | if (!dm_table_supports_discards(t)) { |
5d47c89f MS |
1945 | q->limits.max_discard_sectors = 0; |
1946 | q->limits.max_hw_discard_sectors = 0; | |
1947 | q->limits.discard_granularity = 0; | |
1948 | q->limits.discard_alignment = 0; | |
1949 | q->limits.discard_misaligned = 0; | |
70200574 | 1950 | } |
5ae89a87 | 1951 | |
44abff2c CH |
1952 | if (!dm_table_supports_secure_erase(t)) |
1953 | q->limits.max_secure_erase_sectors = 0; | |
00716545 | 1954 | |
c888a8f9 | 1955 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) { |
519a7e16 | 1956 | wc = true; |
c888a8f9 | 1957 | if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA))) |
519a7e16 | 1958 | fua = true; |
ed8b752b | 1959 | } |
519a7e16 | 1960 | blk_queue_write_cache(q, wc, fua); |
ed8b752b | 1961 | |
7b0800d0 | 1962 | if (dm_table_supports_dax(t, device_not_dax_capable)) { |
8b904b5b | 1963 | blk_queue_flag_set(QUEUE_FLAG_DAX, q); |
7b0800d0 | 1964 | if (dm_table_supports_dax(t, device_not_dax_synchronous_capable)) |
2e9ee095 | 1965 | set_dax_synchronous(t->md->dax_dev); |
03b18887 | 1966 | } else |
dbc62659 RZ |
1967 | blk_queue_flag_clear(QUEUE_FLAG_DAX, q); |
1968 | ||
24f6b603 | 1969 | if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL)) |
273752c9 VG |
1970 | dax_write_cache(t->md->dax_dev, true); |
1971 | ||
c3c4555e | 1972 | /* Ensure that all underlying devices are non-rotational. */ |
24f6b603 | 1973 | if (dm_table_any_dev_attr(t, device_is_rotational, NULL)) |
8b904b5b | 1974 | blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); |
a4c8dd9c JX |
1975 | else |
1976 | blk_queue_flag_set(QUEUE_FLAG_NONROT, q); | |
4693c966 | 1977 | |
ac62d620 CH |
1978 | if (!dm_table_supports_write_zeroes(t)) |
1979 | q->limits.max_write_zeroes_sectors = 0; | |
c1a94672 | 1980 | |
25520d55 | 1981 | dm_table_verify_integrity(t); |
e6ee8c0b | 1982 | |
eb40c0ac ID |
1983 | /* |
1984 | * Some devices don't use blk_integrity but still want stable pages | |
1985 | * because they do their own checksumming. | |
a4c8dd9c JX |
1986 | * If any underlying device requires stable pages, a table must require |
1987 | * them as well. Only targets that support iterate_devices are considered: | |
1988 | * don't want error, zero, etc to require stable pages. | |
eb40c0ac | 1989 | */ |
24f6b603 | 1990 | if (dm_table_any_dev_attr(t, device_requires_stable_pages, NULL)) |
1cb039f3 | 1991 | blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); |
eb40c0ac | 1992 | else |
1cb039f3 | 1993 | blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q); |
eb40c0ac | 1994 | |
c3c4555e MB |
1995 | /* |
1996 | * Determine whether or not this queue's I/O timings contribute | |
1997 | * to the entropy pool, Only request-based targets use this. | |
1998 | * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not | |
1999 | * have it set. | |
2000 | */ | |
24f6b603 JX |
2001 | if (blk_queue_add_random(q) && |
2002 | dm_table_any_dev_attr(t, device_is_not_random, NULL)) | |
8b904b5b | 2003 | blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); |
bf505456 | 2004 | |
bb37d772 DLM |
2005 | /* |
2006 | * For a zoned target, setup the zones related queue attributes | |
2007 | * and resources necessary for zone append emulation if necessary. | |
2008 | */ | |
2009 | if (blk_queue_is_zoned(q)) { | |
2010 | r = dm_set_zones_restrictions(t, q); | |
2011 | if (r) | |
2012 | return r; | |
442761fd MS |
2013 | if (!static_key_enabled(&zoned_enabled.key)) |
2014 | static_branch_enable(&zoned_enabled); | |
bb37d772 | 2015 | } |
c6d6e9b0 | 2016 | |
cb77cb5a | 2017 | dm_update_crypto_profile(q, t); |
471aa704 | 2018 | disk_update_readahead(t->md->disk); |
bb37d772 | 2019 | |
b99fdcdc ML |
2020 | /* |
2021 | * Check for request-based device is left to | |
2022 | * dm_mq_init_request_queue()->blk_mq_init_allocated_queue(). | |
2023 | * | |
2024 | * For bio-based device, only set QUEUE_FLAG_POLL when all | |
2025 | * underlying devices supporting polling. | |
2026 | */ | |
2027 | if (__table_type_bio_based(t->type)) { | |
2028 | if (dm_table_supports_poll(t)) | |
2029 | blk_queue_flag_set(QUEUE_FLAG_POLL, q); | |
2030 | else | |
2031 | blk_queue_flag_clear(QUEUE_FLAG_POLL, q); | |
2032 | } | |
2033 | ||
bb37d772 | 2034 | return 0; |
1da177e4 LT |
2035 | } |
2036 | ||
1da177e4 LT |
2037 | struct list_head *dm_table_get_devices(struct dm_table *t) |
2038 | { | |
2039 | return &t->devices; | |
2040 | } | |
2041 | ||
aeb5d727 | 2042 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
2043 | { |
2044 | return t->mode; | |
2045 | } | |
08649012 | 2046 | EXPORT_SYMBOL(dm_table_get_mode); |
1da177e4 | 2047 | |
d67ee213 MS |
2048 | enum suspend_mode { |
2049 | PRESUSPEND, | |
2050 | PRESUSPEND_UNDO, | |
2051 | POSTSUSPEND, | |
2052 | }; | |
2053 | ||
2054 | static void suspend_targets(struct dm_table *t, enum suspend_mode mode) | |
1da177e4 | 2055 | { |
1ea0654e BVA |
2056 | lockdep_assert_held(&t->md->suspend_lock); |
2057 | ||
564b5c54 MS |
2058 | for (unsigned int i = 0; i < t->num_targets; i++) { |
2059 | struct dm_target *ti = dm_table_get_target(t, i); | |
2060 | ||
d67ee213 MS |
2061 | switch (mode) { |
2062 | case PRESUSPEND: | |
2063 | if (ti->type->presuspend) | |
2064 | ti->type->presuspend(ti); | |
2065 | break; | |
2066 | case PRESUSPEND_UNDO: | |
2067 | if (ti->type->presuspend_undo) | |
2068 | ti->type->presuspend_undo(ti); | |
2069 | break; | |
2070 | case POSTSUSPEND: | |
1da177e4 LT |
2071 | if (ti->type->postsuspend) |
2072 | ti->type->postsuspend(ti); | |
d67ee213 MS |
2073 | break; |
2074 | } | |
1da177e4 LT |
2075 | } |
2076 | } | |
2077 | ||
2078 | void dm_table_presuspend_targets(struct dm_table *t) | |
2079 | { | |
cf222b37 AK |
2080 | if (!t) |
2081 | return; | |
2082 | ||
d67ee213 MS |
2083 | suspend_targets(t, PRESUSPEND); |
2084 | } | |
2085 | ||
2086 | void dm_table_presuspend_undo_targets(struct dm_table *t) | |
2087 | { | |
2088 | if (!t) | |
2089 | return; | |
2090 | ||
2091 | suspend_targets(t, PRESUSPEND_UNDO); | |
1da177e4 LT |
2092 | } |
2093 | ||
2094 | void dm_table_postsuspend_targets(struct dm_table *t) | |
2095 | { | |
cf222b37 AK |
2096 | if (!t) |
2097 | return; | |
2098 | ||
d67ee213 | 2099 | suspend_targets(t, POSTSUSPEND); |
1da177e4 LT |
2100 | } |
2101 | ||
8757b776 | 2102 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 2103 | { |
564b5c54 MS |
2104 | unsigned int i; |
2105 | int r = 0; | |
8757b776 | 2106 | |
1ea0654e BVA |
2107 | lockdep_assert_held(&t->md->suspend_lock); |
2108 | ||
8757b776 | 2109 | for (i = 0; i < t->num_targets; i++) { |
564b5c54 | 2110 | struct dm_target *ti = dm_table_get_target(t, i); |
8757b776 MB |
2111 | |
2112 | if (!ti->type->preresume) | |
2113 | continue; | |
2114 | ||
2115 | r = ti->type->preresume(ti); | |
7833b08e MS |
2116 | if (r) { |
2117 | DMERR("%s: %s: preresume failed, error = %d", | |
2118 | dm_device_name(t->md), ti->type->name, r); | |
8757b776 | 2119 | return r; |
7833b08e | 2120 | } |
8757b776 | 2121 | } |
1da177e4 LT |
2122 | |
2123 | for (i = 0; i < t->num_targets; i++) { | |
564b5c54 | 2124 | struct dm_target *ti = dm_table_get_target(t, i); |
1da177e4 LT |
2125 | |
2126 | if (ti->type->resume) | |
2127 | ti->type->resume(ti); | |
2128 | } | |
8757b776 MB |
2129 | |
2130 | return 0; | |
1da177e4 LT |
2131 | } |
2132 | ||
1134e5ae MA |
2133 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
2134 | { | |
1134e5ae MA |
2135 | return t->md; |
2136 | } | |
08649012 | 2137 | EXPORT_SYMBOL(dm_table_get_md); |
1134e5ae | 2138 | |
f349b0a3 MM |
2139 | const char *dm_table_device_name(struct dm_table *t) |
2140 | { | |
2141 | return dm_device_name(t->md); | |
2142 | } | |
2143 | EXPORT_SYMBOL_GPL(dm_table_device_name); | |
2144 | ||
9974fa2c MS |
2145 | void dm_table_run_md_queue_async(struct dm_table *t) |
2146 | { | |
9974fa2c MS |
2147 | if (!dm_table_request_based(t)) |
2148 | return; | |
2149 | ||
33bd6f06 MS |
2150 | if (t->md->queue) |
2151 | blk_mq_run_hw_queues(t->md->queue, true); | |
9974fa2c MS |
2152 | } |
2153 | EXPORT_SYMBOL(dm_table_run_md_queue_async); | |
2154 |