Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (C) 2001 Sistina Software (UK) Limited. | |
d5816876 | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
1da177e4 LT |
4 | * |
5 | * This file is released under the GPL. | |
6 | */ | |
7 | ||
8 | #include "dm.h" | |
9 | ||
10 | #include <linux/module.h> | |
11 | #include <linux/vmalloc.h> | |
12 | #include <linux/blkdev.h> | |
13 | #include <linux/namei.h> | |
14 | #include <linux/ctype.h> | |
e7d2860b | 15 | #include <linux/string.h> |
1da177e4 LT |
16 | #include <linux/slab.h> |
17 | #include <linux/interrupt.h> | |
48c9c27b | 18 | #include <linux/mutex.h> |
d5816876 | 19 | #include <linux/delay.h> |
60063497 | 20 | #include <linux/atomic.h> |
1da177e4 | 21 | |
72d94861 AK |
22 | #define DM_MSG_PREFIX "table" |
23 | ||
1da177e4 LT |
24 | #define MAX_DEPTH 16 |
25 | #define NODE_SIZE L1_CACHE_BYTES | |
26 | #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t)) | |
27 | #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1) | |
28 | ||
29 | struct dm_table { | |
1134e5ae | 30 | struct mapped_device *md; |
e6ee8c0b | 31 | unsigned type; |
1da177e4 LT |
32 | |
33 | /* btree table */ | |
34 | unsigned int depth; | |
35 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | |
36 | sector_t *index[MAX_DEPTH]; | |
37 | ||
38 | unsigned int num_targets; | |
39 | unsigned int num_allocated; | |
40 | sector_t *highs; | |
41 | struct dm_target *targets; | |
42 | ||
36a0456f | 43 | struct target_type *immutable_target_type; |
a91a2785 | 44 | unsigned integrity_supported:1; |
3791e2fc | 45 | unsigned singleton:1; |
5ae89a87 | 46 | |
1da177e4 LT |
47 | /* |
48 | * Indicates the rw permissions for the new logical | |
49 | * device. This should be a combination of FMODE_READ | |
50 | * and FMODE_WRITE. | |
51 | */ | |
aeb5d727 | 52 | fmode_t mode; |
1da177e4 LT |
53 | |
54 | /* a list of devices used by this table */ | |
55 | struct list_head devices; | |
56 | ||
1da177e4 LT |
57 | /* events get handed up using this callback */ |
58 | void (*event_fn)(void *); | |
59 | void *event_context; | |
e6ee8c0b KU |
60 | |
61 | struct dm_md_mempools *mempools; | |
9d357b07 N |
62 | |
63 | struct list_head target_callbacks; | |
1da177e4 LT |
64 | }; |
65 | ||
66 | /* | |
67 | * Similar to ceiling(log_size(n)) | |
68 | */ | |
69 | static unsigned int int_log(unsigned int n, unsigned int base) | |
70 | { | |
71 | int result = 0; | |
72 | ||
73 | while (n > 1) { | |
74 | n = dm_div_up(n, base); | |
75 | result++; | |
76 | } | |
77 | ||
78 | return result; | |
79 | } | |
80 | ||
1da177e4 LT |
81 | /* |
82 | * Calculate the index of the child node of the n'th node k'th key. | |
83 | */ | |
84 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
85 | { | |
86 | return (n * CHILDREN_PER_NODE) + k; | |
87 | } | |
88 | ||
89 | /* | |
90 | * Return the n'th node of level l from table t. | |
91 | */ | |
92 | static inline sector_t *get_node(struct dm_table *t, | |
93 | unsigned int l, unsigned int n) | |
94 | { | |
95 | return t->index[l] + (n * KEYS_PER_NODE); | |
96 | } | |
97 | ||
98 | /* | |
99 | * Return the highest key that you could lookup from the n'th | |
100 | * node on level l of the btree. | |
101 | */ | |
102 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
103 | { | |
104 | for (; l < t->depth - 1; l++) | |
105 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
106 | ||
107 | if (n >= t->counts[l]) | |
108 | return (sector_t) - 1; | |
109 | ||
110 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
111 | } | |
112 | ||
113 | /* | |
114 | * Fills in a level of the btree based on the highs of the level | |
115 | * below it. | |
116 | */ | |
117 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
118 | { | |
119 | unsigned int n, k; | |
120 | sector_t *node; | |
121 | ||
122 | for (n = 0U; n < t->counts[l]; n++) { | |
123 | node = get_node(t, l, n); | |
124 | ||
125 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
126 | node[k] = high(t, l + 1, get_child(n, k)); | |
127 | } | |
128 | ||
129 | return 0; | |
130 | } | |
131 | ||
132 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
133 | { | |
134 | unsigned long size; | |
135 | void *addr; | |
136 | ||
137 | /* | |
138 | * Check that we're not going to overflow. | |
139 | */ | |
140 | if (nmemb > (ULONG_MAX / elem_size)) | |
141 | return NULL; | |
142 | ||
143 | size = nmemb * elem_size; | |
e29e65aa | 144 | addr = vzalloc(size); |
1da177e4 LT |
145 | |
146 | return addr; | |
147 | } | |
08649012 | 148 | EXPORT_SYMBOL(dm_vcalloc); |
1da177e4 LT |
149 | |
150 | /* | |
151 | * highs, and targets are managed as dynamic arrays during a | |
152 | * table load. | |
153 | */ | |
154 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
155 | { | |
156 | sector_t *n_highs; | |
157 | struct dm_target *n_targets; | |
1da177e4 LT |
158 | |
159 | /* | |
160 | * Allocate both the target array and offset array at once. | |
512875bd JN |
161 | * Append an empty entry to catch sectors beyond the end of |
162 | * the device. | |
1da177e4 | 163 | */ |
512875bd | 164 | n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) + |
1da177e4 LT |
165 | sizeof(sector_t)); |
166 | if (!n_highs) | |
167 | return -ENOMEM; | |
168 | ||
169 | n_targets = (struct dm_target *) (n_highs + num); | |
170 | ||
57a2f238 | 171 | memset(n_highs, -1, sizeof(*n_highs) * num); |
1da177e4 LT |
172 | vfree(t->highs); |
173 | ||
174 | t->num_allocated = num; | |
175 | t->highs = n_highs; | |
176 | t->targets = n_targets; | |
177 | ||
178 | return 0; | |
179 | } | |
180 | ||
aeb5d727 | 181 | int dm_table_create(struct dm_table **result, fmode_t mode, |
1134e5ae | 182 | unsigned num_targets, struct mapped_device *md) |
1da177e4 | 183 | { |
094262db | 184 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
185 | |
186 | if (!t) | |
187 | return -ENOMEM; | |
188 | ||
1da177e4 | 189 | INIT_LIST_HEAD(&t->devices); |
9d357b07 | 190 | INIT_LIST_HEAD(&t->target_callbacks); |
1da177e4 LT |
191 | |
192 | if (!num_targets) | |
193 | num_targets = KEYS_PER_NODE; | |
194 | ||
195 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
196 | ||
5b2d0657 MP |
197 | if (!num_targets) { |
198 | kfree(t); | |
199 | return -ENOMEM; | |
200 | } | |
201 | ||
1da177e4 LT |
202 | if (alloc_targets(t, num_targets)) { |
203 | kfree(t); | |
1da177e4 LT |
204 | return -ENOMEM; |
205 | } | |
206 | ||
207 | t->mode = mode; | |
1134e5ae | 208 | t->md = md; |
1da177e4 LT |
209 | *result = t; |
210 | return 0; | |
211 | } | |
212 | ||
213 | static void free_devices(struct list_head *devices) | |
214 | { | |
215 | struct list_head *tmp, *next; | |
216 | ||
afb24528 | 217 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
218 | struct dm_dev_internal *dd = |
219 | list_entry(tmp, struct dm_dev_internal, list); | |
1b6da754 JB |
220 | DMWARN("dm_table_destroy: dm_put_device call missing for %s", |
221 | dd->dm_dev.name); | |
1da177e4 LT |
222 | kfree(dd); |
223 | } | |
224 | } | |
225 | ||
d5816876 | 226 | void dm_table_destroy(struct dm_table *t) |
1da177e4 LT |
227 | { |
228 | unsigned int i; | |
229 | ||
a7940155 AK |
230 | if (!t) |
231 | return; | |
232 | ||
26803b9f | 233 | /* free the indexes */ |
1da177e4 LT |
234 | if (t->depth >= 2) |
235 | vfree(t->index[t->depth - 2]); | |
236 | ||
237 | /* free the targets */ | |
238 | for (i = 0; i < t->num_targets; i++) { | |
239 | struct dm_target *tgt = t->targets + i; | |
240 | ||
241 | if (tgt->type->dtr) | |
242 | tgt->type->dtr(tgt); | |
243 | ||
244 | dm_put_target_type(tgt->type); | |
245 | } | |
246 | ||
247 | vfree(t->highs); | |
248 | ||
249 | /* free the device list */ | |
574ce07e | 250 | free_devices(&t->devices); |
1da177e4 | 251 | |
e6ee8c0b KU |
252 | dm_free_md_mempools(t->mempools); |
253 | ||
1da177e4 LT |
254 | kfree(t); |
255 | } | |
256 | ||
1da177e4 LT |
257 | /* |
258 | * See if we've already got a device in the list. | |
259 | */ | |
82b1519b | 260 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 261 | { |
82b1519b | 262 | struct dm_dev_internal *dd; |
1da177e4 LT |
263 | |
264 | list_for_each_entry (dd, l, list) | |
82b1519b | 265 | if (dd->dm_dev.bdev->bd_dev == dev) |
1da177e4 LT |
266 | return dd; |
267 | ||
268 | return NULL; | |
269 | } | |
270 | ||
271 | /* | |
272 | * Open a device so we can use it as a map destination. | |
273 | */ | |
82b1519b MP |
274 | static int open_dev(struct dm_dev_internal *d, dev_t dev, |
275 | struct mapped_device *md) | |
1da177e4 LT |
276 | { |
277 | static char *_claim_ptr = "I belong to device-mapper"; | |
278 | struct block_device *bdev; | |
279 | ||
280 | int r; | |
281 | ||
82b1519b | 282 | BUG_ON(d->dm_dev.bdev); |
1da177e4 | 283 | |
d4d77629 | 284 | bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr); |
1da177e4 LT |
285 | if (IS_ERR(bdev)) |
286 | return PTR_ERR(bdev); | |
e09b457b | 287 | |
e09b457b TH |
288 | r = bd_link_disk_holder(bdev, dm_disk(md)); |
289 | if (r) { | |
e525fd89 | 290 | blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL); |
e09b457b TH |
291 | return r; |
292 | } | |
293 | ||
294 | d->dm_dev.bdev = bdev; | |
295 | return 0; | |
1da177e4 LT |
296 | } |
297 | ||
298 | /* | |
299 | * Close a device that we've been using. | |
300 | */ | |
82b1519b | 301 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 302 | { |
82b1519b | 303 | if (!d->dm_dev.bdev) |
1da177e4 LT |
304 | return; |
305 | ||
49731baa | 306 | bd_unlink_disk_holder(d->dm_dev.bdev, dm_disk(md)); |
e525fd89 | 307 | blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL); |
82b1519b | 308 | d->dm_dev.bdev = NULL; |
1da177e4 LT |
309 | } |
310 | ||
311 | /* | |
f6a1ed10 | 312 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 313 | */ |
f6a1ed10 MP |
314 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
315 | sector_t start, sector_t len, void *data) | |
1da177e4 | 316 | { |
f4808ca9 | 317 | struct request_queue *q; |
754c5fc7 MS |
318 | struct queue_limits *limits = data; |
319 | struct block_device *bdev = dev->bdev; | |
320 | sector_t dev_size = | |
321 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 322 | unsigned short logical_block_size_sectors = |
754c5fc7 | 323 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 324 | char b[BDEVNAME_SIZE]; |
2cd54d9b | 325 | |
f4808ca9 MB |
326 | /* |
327 | * Some devices exist without request functions, | |
328 | * such as loop devices not yet bound to backing files. | |
329 | * Forbid the use of such devices. | |
330 | */ | |
331 | q = bdev_get_queue(bdev); | |
332 | if (!q || !q->make_request_fn) { | |
333 | DMWARN("%s: %s is not yet initialised: " | |
334 | "start=%llu, len=%llu, dev_size=%llu", | |
335 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
336 | (unsigned long long)start, | |
337 | (unsigned long long)len, | |
338 | (unsigned long long)dev_size); | |
339 | return 1; | |
340 | } | |
341 | ||
2cd54d9b | 342 | if (!dev_size) |
f6a1ed10 | 343 | return 0; |
2cd54d9b | 344 | |
5dea271b | 345 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
346 | DMWARN("%s: %s too small for target: " |
347 | "start=%llu, len=%llu, dev_size=%llu", | |
348 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
349 | (unsigned long long)start, | |
350 | (unsigned long long)len, | |
351 | (unsigned long long)dev_size); | |
f6a1ed10 | 352 | return 1; |
02acc3a4 MS |
353 | } |
354 | ||
355 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 356 | return 0; |
02acc3a4 MS |
357 | |
358 | if (start & (logical_block_size_sectors - 1)) { | |
359 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 360 | "logical block size %u of %s", |
02acc3a4 MS |
361 | dm_device_name(ti->table->md), |
362 | (unsigned long long)start, | |
754c5fc7 | 363 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 364 | return 1; |
02acc3a4 MS |
365 | } |
366 | ||
5dea271b | 367 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 368 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 369 | "logical block size %u of %s", |
02acc3a4 | 370 | dm_device_name(ti->table->md), |
5dea271b | 371 | (unsigned long long)len, |
754c5fc7 | 372 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 373 | return 1; |
02acc3a4 MS |
374 | } |
375 | ||
f6a1ed10 | 376 | return 0; |
1da177e4 LT |
377 | } |
378 | ||
379 | /* | |
570b9d96 | 380 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 381 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
382 | * device and not to touch the existing bdev field in case |
383 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 384 | */ |
aeb5d727 | 385 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 386 | struct mapped_device *md) |
1da177e4 LT |
387 | { |
388 | int r; | |
570b9d96 | 389 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 390 | |
570b9d96 AK |
391 | dd_new = dd_old = *dd; |
392 | ||
393 | dd_new.dm_dev.mode |= new_mode; | |
394 | dd_new.dm_dev.bdev = NULL; | |
395 | ||
396 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
397 | if (r) | |
398 | return r; | |
1da177e4 | 399 | |
82b1519b | 400 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 401 | close_dev(&dd_old, md); |
1da177e4 | 402 | |
570b9d96 | 403 | return 0; |
1da177e4 LT |
404 | } |
405 | ||
406 | /* | |
407 | * Add a device to the list, or just increment the usage count if | |
408 | * it's already present. | |
409 | */ | |
08649012 MS |
410 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
411 | struct dm_dev **result) | |
1da177e4 LT |
412 | { |
413 | int r; | |
69a2ce72 | 414 | dev_t uninitialized_var(dev); |
82b1519b | 415 | struct dm_dev_internal *dd; |
1da177e4 | 416 | unsigned int major, minor; |
08649012 | 417 | struct dm_table *t = ti->table; |
31998ef1 | 418 | char dummy; |
1da177e4 | 419 | |
547bc926 | 420 | BUG_ON(!t); |
1da177e4 | 421 | |
31998ef1 | 422 | if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) { |
1da177e4 LT |
423 | /* Extract the major/minor numbers */ |
424 | dev = MKDEV(major, minor); | |
425 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
426 | return -EOVERFLOW; | |
427 | } else { | |
428 | /* convert the path to a device */ | |
72e8264e CH |
429 | struct block_device *bdev = lookup_bdev(path); |
430 | ||
431 | if (IS_ERR(bdev)) | |
432 | return PTR_ERR(bdev); | |
433 | dev = bdev->bd_dev; | |
434 | bdput(bdev); | |
1da177e4 LT |
435 | } |
436 | ||
437 | dd = find_device(&t->devices, dev); | |
438 | if (!dd) { | |
439 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
440 | if (!dd) | |
441 | return -ENOMEM; | |
442 | ||
82b1519b MP |
443 | dd->dm_dev.mode = mode; |
444 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 445 | |
f165921d | 446 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
447 | kfree(dd); |
448 | return r; | |
449 | } | |
450 | ||
82b1519b | 451 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
452 | |
453 | atomic_set(&dd->count, 0); | |
454 | list_add(&dd->list, &t->devices); | |
455 | ||
82b1519b | 456 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 457 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
458 | if (r) |
459 | return r; | |
460 | } | |
461 | atomic_inc(&dd->count); | |
462 | ||
82b1519b | 463 | *result = &dd->dm_dev; |
1da177e4 LT |
464 | return 0; |
465 | } | |
08649012 | 466 | EXPORT_SYMBOL(dm_get_device); |
1da177e4 | 467 | |
754c5fc7 | 468 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 469 | sector_t start, sector_t len, void *data) |
1da177e4 | 470 | { |
754c5fc7 MS |
471 | struct queue_limits *limits = data; |
472 | struct block_device *bdev = dev->bdev; | |
165125e1 | 473 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
474 | char b[BDEVNAME_SIZE]; |
475 | ||
476 | if (unlikely(!q)) { | |
477 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
478 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 479 | return 0; |
0c2322e4 | 480 | } |
3cb40214 | 481 | |
b27d7f16 MP |
482 | if (bdev_stack_limits(limits, bdev, start) < 0) |
483 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
484 | "physical_block_size=%u, logical_block_size=%u, " |
485 | "alignment_offset=%u, start=%llu", | |
486 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
487 | q->limits.physical_block_size, | |
488 | q->limits.logical_block_size, | |
489 | q->limits.alignment_offset, | |
b27d7f16 | 490 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 491 | |
9980c638 MB |
492 | /* |
493 | * Check if merge fn is supported. | |
494 | * If not we'll force DM to use PAGE_SIZE or | |
495 | * smaller I/O, just to be safe. | |
3cb40214 | 496 | */ |
d5b9dd04 | 497 | if (dm_queue_merge_is_compulsory(q) && !ti->type->merge) |
72d4cd9f MS |
498 | blk_limits_max_hw_sectors(limits, |
499 | (unsigned int) (PAGE_SIZE >> 9)); | |
754c5fc7 | 500 | return 0; |
3cb40214 BR |
501 | } |
502 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 503 | |
1da177e4 | 504 | /* |
08649012 | 505 | * Decrement a device's use count and remove it if necessary. |
1da177e4 | 506 | */ |
82b1519b | 507 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 508 | { |
82b1519b MP |
509 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
510 | dm_dev); | |
511 | ||
1da177e4 | 512 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 513 | close_dev(dd, ti->table->md); |
1da177e4 LT |
514 | list_del(&dd->list); |
515 | kfree(dd); | |
516 | } | |
517 | } | |
08649012 | 518 | EXPORT_SYMBOL(dm_put_device); |
1da177e4 LT |
519 | |
520 | /* | |
521 | * Checks to see if the target joins onto the end of the table. | |
522 | */ | |
523 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
524 | { | |
525 | struct dm_target *prev; | |
526 | ||
527 | if (!table->num_targets) | |
528 | return !ti->begin; | |
529 | ||
530 | prev = &table->targets[table->num_targets - 1]; | |
531 | return (ti->begin == (prev->begin + prev->len)); | |
532 | } | |
533 | ||
534 | /* | |
535 | * Used to dynamically allocate the arg array. | |
f36afb39 MP |
536 | * |
537 | * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must | |
538 | * process messages even if some device is suspended. These messages have a | |
539 | * small fixed number of arguments. | |
540 | * | |
541 | * On the other hand, dm-switch needs to process bulk data using messages and | |
542 | * excessive use of GFP_NOIO could cause trouble. | |
1da177e4 LT |
543 | */ |
544 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
545 | { | |
546 | char **argv; | |
547 | unsigned new_size; | |
f36afb39 | 548 | gfp_t gfp; |
1da177e4 | 549 | |
f36afb39 MP |
550 | if (*array_size) { |
551 | new_size = *array_size * 2; | |
552 | gfp = GFP_KERNEL; | |
553 | } else { | |
554 | new_size = 8; | |
555 | gfp = GFP_NOIO; | |
556 | } | |
557 | argv = kmalloc(new_size * sizeof(*argv), gfp); | |
1da177e4 LT |
558 | if (argv) { |
559 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
560 | *array_size = new_size; | |
561 | } | |
562 | ||
563 | kfree(old_argv); | |
564 | return argv; | |
565 | } | |
566 | ||
567 | /* | |
568 | * Destructively splits up the argument list to pass to ctr. | |
569 | */ | |
570 | int dm_split_args(int *argc, char ***argvp, char *input) | |
571 | { | |
572 | char *start, *end = input, *out, **argv = NULL; | |
573 | unsigned array_size = 0; | |
574 | ||
575 | *argc = 0; | |
814d6862 DT |
576 | |
577 | if (!input) { | |
578 | *argvp = NULL; | |
579 | return 0; | |
580 | } | |
581 | ||
1da177e4 LT |
582 | argv = realloc_argv(&array_size, argv); |
583 | if (!argv) | |
584 | return -ENOMEM; | |
585 | ||
586 | while (1) { | |
1da177e4 | 587 | /* Skip whitespace */ |
e7d2860b | 588 | start = skip_spaces(end); |
1da177e4 LT |
589 | |
590 | if (!*start) | |
591 | break; /* success, we hit the end */ | |
592 | ||
593 | /* 'out' is used to remove any back-quotes */ | |
594 | end = out = start; | |
595 | while (*end) { | |
596 | /* Everything apart from '\0' can be quoted */ | |
597 | if (*end == '\\' && *(end + 1)) { | |
598 | *out++ = *(end + 1); | |
599 | end += 2; | |
600 | continue; | |
601 | } | |
602 | ||
603 | if (isspace(*end)) | |
604 | break; /* end of token */ | |
605 | ||
606 | *out++ = *end++; | |
607 | } | |
608 | ||
609 | /* have we already filled the array ? */ | |
610 | if ((*argc + 1) > array_size) { | |
611 | argv = realloc_argv(&array_size, argv); | |
612 | if (!argv) | |
613 | return -ENOMEM; | |
614 | } | |
615 | ||
616 | /* we know this is whitespace */ | |
617 | if (*end) | |
618 | end++; | |
619 | ||
620 | /* terminate the string and put it in the array */ | |
621 | *out = '\0'; | |
622 | argv[*argc] = start; | |
623 | (*argc)++; | |
624 | } | |
625 | ||
626 | *argvp = argv; | |
627 | return 0; | |
628 | } | |
629 | ||
be6d4305 MS |
630 | /* |
631 | * Impose necessary and sufficient conditions on a devices's table such | |
632 | * that any incoming bio which respects its logical_block_size can be | |
633 | * processed successfully. If it falls across the boundary between | |
634 | * two or more targets, the size of each piece it gets split into must | |
635 | * be compatible with the logical_block_size of the target processing it. | |
636 | */ | |
754c5fc7 MS |
637 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
638 | struct queue_limits *limits) | |
be6d4305 MS |
639 | { |
640 | /* | |
641 | * This function uses arithmetic modulo the logical_block_size | |
642 | * (in units of 512-byte sectors). | |
643 | */ | |
644 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 645 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
646 | |
647 | /* | |
648 | * Offset of the start of the next table entry, mod logical_block_size. | |
649 | */ | |
650 | unsigned short next_target_start = 0; | |
651 | ||
652 | /* | |
653 | * Given an aligned bio that extends beyond the end of a | |
654 | * target, how many sectors must the next target handle? | |
655 | */ | |
656 | unsigned short remaining = 0; | |
657 | ||
658 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 659 | struct queue_limits ti_limits; |
be6d4305 MS |
660 | unsigned i = 0; |
661 | ||
662 | /* | |
663 | * Check each entry in the table in turn. | |
664 | */ | |
665 | while (i < dm_table_get_num_targets(table)) { | |
666 | ti = dm_table_get_target(table, i++); | |
667 | ||
b1bd055d | 668 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
669 | |
670 | /* combine all target devices' limits */ | |
671 | if (ti->type->iterate_devices) | |
672 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
673 | &ti_limits); | |
674 | ||
be6d4305 MS |
675 | /* |
676 | * If the remaining sectors fall entirely within this | |
677 | * table entry are they compatible with its logical_block_size? | |
678 | */ | |
679 | if (remaining < ti->len && | |
754c5fc7 | 680 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
681 | SECTOR_SHIFT) - 1)) |
682 | break; /* Error */ | |
683 | ||
684 | next_target_start = | |
685 | (unsigned short) ((next_target_start + ti->len) & | |
686 | (device_logical_block_size_sects - 1)); | |
687 | remaining = next_target_start ? | |
688 | device_logical_block_size_sects - next_target_start : 0; | |
689 | } | |
690 | ||
691 | if (remaining) { | |
692 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 693 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
694 | dm_device_name(table->md), i, |
695 | (unsigned long long) ti->begin, | |
696 | (unsigned long long) ti->len, | |
754c5fc7 | 697 | limits->logical_block_size); |
be6d4305 MS |
698 | return -EINVAL; |
699 | } | |
700 | ||
701 | return 0; | |
702 | } | |
703 | ||
1da177e4 LT |
704 | int dm_table_add_target(struct dm_table *t, const char *type, |
705 | sector_t start, sector_t len, char *params) | |
706 | { | |
707 | int r = -EINVAL, argc; | |
708 | char **argv; | |
709 | struct dm_target *tgt; | |
710 | ||
3791e2fc AK |
711 | if (t->singleton) { |
712 | DMERR("%s: target type %s must appear alone in table", | |
713 | dm_device_name(t->md), t->targets->type->name); | |
714 | return -EINVAL; | |
715 | } | |
716 | ||
57a2f238 | 717 | BUG_ON(t->num_targets >= t->num_allocated); |
1da177e4 LT |
718 | |
719 | tgt = t->targets + t->num_targets; | |
720 | memset(tgt, 0, sizeof(*tgt)); | |
721 | ||
722 | if (!len) { | |
72d94861 | 723 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
724 | return -EINVAL; |
725 | } | |
726 | ||
727 | tgt->type = dm_get_target_type(type); | |
728 | if (!tgt->type) { | |
72d94861 AK |
729 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
730 | type); | |
1da177e4 LT |
731 | return -EINVAL; |
732 | } | |
733 | ||
3791e2fc AK |
734 | if (dm_target_needs_singleton(tgt->type)) { |
735 | if (t->num_targets) { | |
736 | DMERR("%s: target type %s must appear alone in table", | |
737 | dm_device_name(t->md), type); | |
738 | return -EINVAL; | |
739 | } | |
740 | t->singleton = 1; | |
741 | } | |
742 | ||
cc6cbe14 AK |
743 | if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) { |
744 | DMERR("%s: target type %s may not be included in read-only tables", | |
745 | dm_device_name(t->md), type); | |
746 | return -EINVAL; | |
747 | } | |
748 | ||
36a0456f AK |
749 | if (t->immutable_target_type) { |
750 | if (t->immutable_target_type != tgt->type) { | |
751 | DMERR("%s: immutable target type %s cannot be mixed with other target types", | |
752 | dm_device_name(t->md), t->immutable_target_type->name); | |
753 | return -EINVAL; | |
754 | } | |
755 | } else if (dm_target_is_immutable(tgt->type)) { | |
756 | if (t->num_targets) { | |
757 | DMERR("%s: immutable target type %s cannot be mixed with other target types", | |
758 | dm_device_name(t->md), tgt->type->name); | |
759 | return -EINVAL; | |
760 | } | |
761 | t->immutable_target_type = tgt->type; | |
762 | } | |
763 | ||
1da177e4 LT |
764 | tgt->table = t; |
765 | tgt->begin = start; | |
766 | tgt->len = len; | |
767 | tgt->error = "Unknown error"; | |
768 | ||
769 | /* | |
770 | * Does this target adjoin the previous one ? | |
771 | */ | |
772 | if (!adjoin(t, tgt)) { | |
773 | tgt->error = "Gap in table"; | |
774 | r = -EINVAL; | |
775 | goto bad; | |
776 | } | |
777 | ||
778 | r = dm_split_args(&argc, &argv, params); | |
779 | if (r) { | |
780 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
781 | goto bad; | |
782 | } | |
783 | ||
784 | r = tgt->type->ctr(tgt, argc, argv); | |
785 | kfree(argv); | |
786 | if (r) | |
787 | goto bad; | |
788 | ||
789 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
790 | ||
55a62eef AK |
791 | if (!tgt->num_discard_bios && tgt->discards_supported) |
792 | DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.", | |
936688d7 | 793 | dm_device_name(t->md), type); |
5ae89a87 | 794 | |
1da177e4 LT |
795 | return 0; |
796 | ||
797 | bad: | |
72d94861 | 798 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
799 | dm_put_target_type(tgt->type); |
800 | return r; | |
801 | } | |
802 | ||
498f0103 MS |
803 | /* |
804 | * Target argument parsing helpers. | |
805 | */ | |
806 | static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
807 | unsigned *value, char **error, unsigned grouped) | |
808 | { | |
809 | const char *arg_str = dm_shift_arg(arg_set); | |
31998ef1 | 810 | char dummy; |
498f0103 MS |
811 | |
812 | if (!arg_str || | |
31998ef1 | 813 | (sscanf(arg_str, "%u%c", value, &dummy) != 1) || |
498f0103 MS |
814 | (*value < arg->min) || |
815 | (*value > arg->max) || | |
816 | (grouped && arg_set->argc < *value)) { | |
817 | *error = arg->error; | |
818 | return -EINVAL; | |
819 | } | |
820 | ||
821 | return 0; | |
822 | } | |
823 | ||
824 | int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
825 | unsigned *value, char **error) | |
826 | { | |
827 | return validate_next_arg(arg, arg_set, value, error, 0); | |
828 | } | |
829 | EXPORT_SYMBOL(dm_read_arg); | |
830 | ||
831 | int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
832 | unsigned *value, char **error) | |
833 | { | |
834 | return validate_next_arg(arg, arg_set, value, error, 1); | |
835 | } | |
836 | EXPORT_SYMBOL(dm_read_arg_group); | |
837 | ||
838 | const char *dm_shift_arg(struct dm_arg_set *as) | |
839 | { | |
840 | char *r; | |
841 | ||
842 | if (as->argc) { | |
843 | as->argc--; | |
844 | r = *as->argv; | |
845 | as->argv++; | |
846 | return r; | |
847 | } | |
848 | ||
849 | return NULL; | |
850 | } | |
851 | EXPORT_SYMBOL(dm_shift_arg); | |
852 | ||
853 | void dm_consume_args(struct dm_arg_set *as, unsigned num_args) | |
854 | { | |
855 | BUG_ON(as->argc < num_args); | |
856 | as->argc -= num_args; | |
857 | as->argv += num_args; | |
858 | } | |
859 | EXPORT_SYMBOL(dm_consume_args); | |
860 | ||
26803b9f | 861 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
862 | { |
863 | unsigned i; | |
169e2cc2 | 864 | unsigned bio_based = 0, request_based = 0, hybrid = 0; |
e6ee8c0b KU |
865 | struct dm_target *tgt; |
866 | struct dm_dev_internal *dd; | |
867 | struct list_head *devices; | |
169e2cc2 | 868 | unsigned live_md_type; |
e6ee8c0b KU |
869 | |
870 | for (i = 0; i < t->num_targets; i++) { | |
871 | tgt = t->targets + i; | |
169e2cc2 MS |
872 | if (dm_target_hybrid(tgt)) |
873 | hybrid = 1; | |
874 | else if (dm_target_request_based(tgt)) | |
e6ee8c0b KU |
875 | request_based = 1; |
876 | else | |
877 | bio_based = 1; | |
878 | ||
879 | if (bio_based && request_based) { | |
880 | DMWARN("Inconsistent table: different target types" | |
881 | " can't be mixed up"); | |
882 | return -EINVAL; | |
883 | } | |
884 | } | |
885 | ||
169e2cc2 MS |
886 | if (hybrid && !bio_based && !request_based) { |
887 | /* | |
888 | * The targets can work either way. | |
889 | * Determine the type from the live device. | |
890 | * Default to bio-based if device is new. | |
891 | */ | |
169e2cc2 | 892 | live_md_type = dm_get_md_type(t->md); |
169e2cc2 MS |
893 | if (live_md_type == DM_TYPE_REQUEST_BASED) |
894 | request_based = 1; | |
895 | else | |
896 | bio_based = 1; | |
897 | } | |
898 | ||
e6ee8c0b KU |
899 | if (bio_based) { |
900 | /* We must use this table as bio-based */ | |
901 | t->type = DM_TYPE_BIO_BASED; | |
902 | return 0; | |
903 | } | |
904 | ||
905 | BUG_ON(!request_based); /* No targets in this table */ | |
906 | ||
907 | /* Non-request-stackable devices can't be used for request-based dm */ | |
908 | devices = dm_table_get_devices(t); | |
909 | list_for_each_entry(dd, devices, list) { | |
910 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
911 | DMWARN("table load rejected: including" | |
912 | " non-request-stackable devices"); | |
913 | return -EINVAL; | |
914 | } | |
915 | } | |
916 | ||
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) { | |
924 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
925 | return -EINVAL; | |
926 | } | |
927 | ||
928 | t->type = DM_TYPE_REQUEST_BASED; | |
929 | ||
930 | return 0; | |
931 | } | |
932 | ||
933 | unsigned dm_table_get_type(struct dm_table *t) | |
934 | { | |
935 | return t->type; | |
936 | } | |
937 | ||
36a0456f AK |
938 | struct target_type *dm_table_get_immutable_target_type(struct dm_table *t) |
939 | { | |
940 | return t->immutable_target_type; | |
941 | } | |
942 | ||
e6ee8c0b KU |
943 | bool dm_table_request_based(struct dm_table *t) |
944 | { | |
945 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
946 | } | |
947 | ||
473c36df | 948 | static int dm_table_alloc_md_mempools(struct dm_table *t) |
e6ee8c0b KU |
949 | { |
950 | unsigned type = dm_table_get_type(t); | |
c0820cf5 MP |
951 | unsigned per_bio_data_size = 0; |
952 | struct dm_target *tgt; | |
953 | unsigned i; | |
e6ee8c0b KU |
954 | |
955 | if (unlikely(type == DM_TYPE_NONE)) { | |
956 | DMWARN("no table type is set, can't allocate mempools"); | |
957 | return -EINVAL; | |
958 | } | |
959 | ||
c0820cf5 MP |
960 | if (type == DM_TYPE_BIO_BASED) |
961 | for (i = 0; i < t->num_targets; i++) { | |
962 | tgt = t->targets + i; | |
963 | per_bio_data_size = max(per_bio_data_size, tgt->per_bio_data_size); | |
964 | } | |
965 | ||
966 | t->mempools = dm_alloc_md_mempools(type, t->integrity_supported, per_bio_data_size); | |
e6ee8c0b KU |
967 | if (!t->mempools) |
968 | return -ENOMEM; | |
969 | ||
970 | return 0; | |
971 | } | |
972 | ||
973 | void dm_table_free_md_mempools(struct dm_table *t) | |
974 | { | |
975 | dm_free_md_mempools(t->mempools); | |
976 | t->mempools = NULL; | |
977 | } | |
978 | ||
979 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
980 | { | |
981 | return t->mempools; | |
982 | } | |
983 | ||
1da177e4 LT |
984 | static int setup_indexes(struct dm_table *t) |
985 | { | |
986 | int i; | |
987 | unsigned int total = 0; | |
988 | sector_t *indexes; | |
989 | ||
990 | /* allocate the space for *all* the indexes */ | |
991 | for (i = t->depth - 2; i >= 0; i--) { | |
992 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
993 | total += t->counts[i]; | |
994 | } | |
995 | ||
996 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
997 | if (!indexes) | |
998 | return -ENOMEM; | |
999 | ||
1000 | /* set up internal nodes, bottom-up */ | |
82d601dc | 1001 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
1002 | t->index[i] = indexes; |
1003 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
1004 | setup_btree_index(i, t); | |
1005 | } | |
1006 | ||
1007 | return 0; | |
1008 | } | |
1009 | ||
1010 | /* | |
1011 | * Builds the btree to index the map. | |
1012 | */ | |
26803b9f | 1013 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
1014 | { |
1015 | int r = 0; | |
1016 | unsigned int leaf_nodes; | |
1017 | ||
1da177e4 LT |
1018 | /* how many indexes will the btree have ? */ |
1019 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
1020 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
1021 | ||
1022 | /* leaf layer has already been set up */ | |
1023 | t->counts[t->depth - 1] = leaf_nodes; | |
1024 | t->index[t->depth - 1] = t->highs; | |
1025 | ||
1026 | if (t->depth >= 2) | |
1027 | r = setup_indexes(t); | |
1028 | ||
1029 | return r; | |
1030 | } | |
1031 | ||
a63a5cf8 MS |
1032 | /* |
1033 | * Get a disk whose integrity profile reflects the table's profile. | |
1034 | * If %match_all is true, all devices' profiles must match. | |
1035 | * If %match_all is false, all devices must at least have an | |
1036 | * allocated integrity profile; but uninitialized is ok. | |
1037 | * Returns NULL if integrity support was inconsistent or unavailable. | |
1038 | */ | |
1039 | static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t, | |
1040 | bool match_all) | |
1041 | { | |
1042 | struct list_head *devices = dm_table_get_devices(t); | |
1043 | struct dm_dev_internal *dd = NULL; | |
1044 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
1045 | ||
1046 | list_for_each_entry(dd, devices, list) { | |
1047 | template_disk = dd->dm_dev.bdev->bd_disk; | |
1048 | if (!blk_get_integrity(template_disk)) | |
1049 | goto no_integrity; | |
1050 | if (!match_all && !blk_integrity_is_initialized(template_disk)) | |
1051 | continue; /* skip uninitialized profiles */ | |
1052 | else if (prev_disk && | |
1053 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
1054 | goto no_integrity; | |
1055 | prev_disk = template_disk; | |
1056 | } | |
1057 | ||
1058 | return template_disk; | |
1059 | ||
1060 | no_integrity: | |
1061 | if (prev_disk) | |
1062 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
1063 | dm_device_name(t->md), | |
1064 | prev_disk->disk_name, | |
1065 | template_disk->disk_name); | |
1066 | return NULL; | |
1067 | } | |
1068 | ||
26803b9f WD |
1069 | /* |
1070 | * Register the mapped device for blk_integrity support if | |
a63a5cf8 MS |
1071 | * the underlying devices have an integrity profile. But all devices |
1072 | * may not have matching profiles (checking all devices isn't reliable | |
1073 | * during table load because this table may use other DM device(s) which | |
1074 | * must be resumed before they will have an initialized integity profile). | |
1075 | * Stacked DM devices force a 2 stage integrity profile validation: | |
1076 | * 1 - during load, validate all initialized integrity profiles match | |
1077 | * 2 - during resume, validate all integrity profiles match | |
26803b9f WD |
1078 | */ |
1079 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
1080 | { | |
a63a5cf8 | 1081 | struct gendisk *template_disk = NULL; |
26803b9f | 1082 | |
a63a5cf8 MS |
1083 | template_disk = dm_table_get_integrity_disk(t, false); |
1084 | if (!template_disk) | |
1085 | return 0; | |
26803b9f | 1086 | |
a63a5cf8 MS |
1087 | if (!blk_integrity_is_initialized(dm_disk(md))) { |
1088 | t->integrity_supported = 1; | |
1089 | return blk_integrity_register(dm_disk(md), NULL); | |
1090 | } | |
1091 | ||
1092 | /* | |
1093 | * If DM device already has an initalized integrity | |
1094 | * profile the new profile should not conflict. | |
1095 | */ | |
1096 | if (blk_integrity_is_initialized(template_disk) && | |
1097 | blk_integrity_compare(dm_disk(md), template_disk) < 0) { | |
1098 | DMWARN("%s: conflict with existing integrity profile: " | |
1099 | "%s profile mismatch", | |
1100 | dm_device_name(t->md), | |
1101 | template_disk->disk_name); | |
1102 | return 1; | |
1103 | } | |
1104 | ||
1105 | /* Preserve existing initialized integrity profile */ | |
1106 | t->integrity_supported = 1; | |
26803b9f WD |
1107 | return 0; |
1108 | } | |
1109 | ||
1110 | /* | |
1111 | * Prepares the table for use by building the indices, | |
1112 | * setting the type, and allocating mempools. | |
1113 | */ | |
1114 | int dm_table_complete(struct dm_table *t) | |
1115 | { | |
1116 | int r; | |
1117 | ||
1118 | r = dm_table_set_type(t); | |
1119 | if (r) { | |
1120 | DMERR("unable to set table type"); | |
1121 | return r; | |
1122 | } | |
1123 | ||
1124 | r = dm_table_build_index(t); | |
1125 | if (r) { | |
1126 | DMERR("unable to build btrees"); | |
1127 | return r; | |
1128 | } | |
1129 | ||
1130 | r = dm_table_prealloc_integrity(t, t->md); | |
1131 | if (r) { | |
1132 | DMERR("could not register integrity profile."); | |
1133 | return r; | |
1134 | } | |
1135 | ||
1136 | r = dm_table_alloc_md_mempools(t); | |
1137 | if (r) | |
1138 | DMERR("unable to allocate mempools"); | |
1139 | ||
1140 | return r; | |
1141 | } | |
1142 | ||
48c9c27b | 1143 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1144 | void dm_table_event_callback(struct dm_table *t, |
1145 | void (*fn)(void *), void *context) | |
1146 | { | |
48c9c27b | 1147 | mutex_lock(&_event_lock); |
1da177e4 LT |
1148 | t->event_fn = fn; |
1149 | t->event_context = context; | |
48c9c27b | 1150 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1151 | } |
1152 | ||
1153 | void dm_table_event(struct dm_table *t) | |
1154 | { | |
1155 | /* | |
1156 | * You can no longer call dm_table_event() from interrupt | |
1157 | * context, use a bottom half instead. | |
1158 | */ | |
1159 | BUG_ON(in_interrupt()); | |
1160 | ||
48c9c27b | 1161 | mutex_lock(&_event_lock); |
1da177e4 LT |
1162 | if (t->event_fn) |
1163 | t->event_fn(t->event_context); | |
48c9c27b | 1164 | mutex_unlock(&_event_lock); |
1da177e4 | 1165 | } |
08649012 | 1166 | EXPORT_SYMBOL(dm_table_event); |
1da177e4 LT |
1167 | |
1168 | sector_t dm_table_get_size(struct dm_table *t) | |
1169 | { | |
1170 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1171 | } | |
08649012 | 1172 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 LT |
1173 | |
1174 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1175 | { | |
14353539 | 1176 | if (index >= t->num_targets) |
1da177e4 LT |
1177 | return NULL; |
1178 | ||
1179 | return t->targets + index; | |
1180 | } | |
1181 | ||
1182 | /* | |
1183 | * Search the btree for the correct target. | |
512875bd JN |
1184 | * |
1185 | * Caller should check returned pointer with dm_target_is_valid() | |
1186 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1187 | */ |
1188 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1189 | { | |
1190 | unsigned int l, n = 0, k = 0; | |
1191 | sector_t *node; | |
1192 | ||
1193 | for (l = 0; l < t->depth; l++) { | |
1194 | n = get_child(n, k); | |
1195 | node = get_node(t, l, n); | |
1196 | ||
1197 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1198 | if (node[k] >= sector) | |
1199 | break; | |
1200 | } | |
1201 | ||
1202 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1203 | } | |
1204 | ||
3ae70656 MS |
1205 | static int count_device(struct dm_target *ti, struct dm_dev *dev, |
1206 | sector_t start, sector_t len, void *data) | |
1207 | { | |
1208 | unsigned *num_devices = data; | |
1209 | ||
1210 | (*num_devices)++; | |
1211 | ||
1212 | return 0; | |
1213 | } | |
1214 | ||
1215 | /* | |
1216 | * Check whether a table has no data devices attached using each | |
1217 | * target's iterate_devices method. | |
1218 | * Returns false if the result is unknown because a target doesn't | |
1219 | * support iterate_devices. | |
1220 | */ | |
1221 | bool dm_table_has_no_data_devices(struct dm_table *table) | |
1222 | { | |
1223 | struct dm_target *uninitialized_var(ti); | |
1224 | unsigned i = 0, num_devices = 0; | |
1225 | ||
1226 | while (i < dm_table_get_num_targets(table)) { | |
1227 | ti = dm_table_get_target(table, i++); | |
1228 | ||
1229 | if (!ti->type->iterate_devices) | |
1230 | return false; | |
1231 | ||
1232 | ti->type->iterate_devices(ti, count_device, &num_devices); | |
1233 | if (num_devices) | |
1234 | return false; | |
1235 | } | |
1236 | ||
1237 | return true; | |
1238 | } | |
1239 | ||
754c5fc7 MS |
1240 | /* |
1241 | * Establish the new table's queue_limits and validate them. | |
1242 | */ | |
1243 | int dm_calculate_queue_limits(struct dm_table *table, | |
1244 | struct queue_limits *limits) | |
1245 | { | |
1246 | struct dm_target *uninitialized_var(ti); | |
1247 | struct queue_limits ti_limits; | |
1248 | unsigned i = 0; | |
1249 | ||
b1bd055d | 1250 | blk_set_stacking_limits(limits); |
754c5fc7 MS |
1251 | |
1252 | while (i < dm_table_get_num_targets(table)) { | |
b1bd055d | 1253 | blk_set_stacking_limits(&ti_limits); |
754c5fc7 MS |
1254 | |
1255 | ti = dm_table_get_target(table, i++); | |
1256 | ||
1257 | if (!ti->type->iterate_devices) | |
1258 | goto combine_limits; | |
1259 | ||
1260 | /* | |
1261 | * Combine queue limits of all the devices this target uses. | |
1262 | */ | |
1263 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1264 | &ti_limits); | |
1265 | ||
40bea431 MS |
1266 | /* Set I/O hints portion of queue limits */ |
1267 | if (ti->type->io_hints) | |
1268 | ti->type->io_hints(ti, &ti_limits); | |
1269 | ||
754c5fc7 MS |
1270 | /* |
1271 | * Check each device area is consistent with the target's | |
1272 | * overall queue limits. | |
1273 | */ | |
f6a1ed10 MP |
1274 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1275 | &ti_limits)) | |
754c5fc7 MS |
1276 | return -EINVAL; |
1277 | ||
1278 | combine_limits: | |
1279 | /* | |
1280 | * Merge this target's queue limits into the overall limits | |
1281 | * for the table. | |
1282 | */ | |
1283 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1284 | DMWARN("%s: adding target device " |
754c5fc7 | 1285 | "(start sect %llu len %llu) " |
b27d7f16 | 1286 | "caused an alignment inconsistency", |
754c5fc7 MS |
1287 | dm_device_name(table->md), |
1288 | (unsigned long long) ti->begin, | |
1289 | (unsigned long long) ti->len); | |
1290 | } | |
1291 | ||
1292 | return validate_hardware_logical_block_alignment(table, limits); | |
1293 | } | |
1294 | ||
9c47008d MP |
1295 | /* |
1296 | * Set the integrity profile for this device if all devices used have | |
a63a5cf8 MS |
1297 | * matching profiles. We're quite deep in the resume path but still |
1298 | * don't know if all devices (particularly DM devices this device | |
1299 | * may be stacked on) have matching profiles. Even if the profiles | |
1300 | * don't match we have no way to fail (to resume) at this point. | |
9c47008d MP |
1301 | */ |
1302 | static void dm_table_set_integrity(struct dm_table *t) | |
1303 | { | |
a63a5cf8 | 1304 | struct gendisk *template_disk = NULL; |
9c47008d MP |
1305 | |
1306 | if (!blk_get_integrity(dm_disk(t->md))) | |
1307 | return; | |
1308 | ||
a63a5cf8 | 1309 | template_disk = dm_table_get_integrity_disk(t, true); |
876fbba1 MS |
1310 | if (template_disk) |
1311 | blk_integrity_register(dm_disk(t->md), | |
1312 | blk_get_integrity(template_disk)); | |
1313 | else if (blk_integrity_is_initialized(dm_disk(t->md))) | |
a63a5cf8 MS |
1314 | DMWARN("%s: device no longer has a valid integrity profile", |
1315 | dm_device_name(t->md)); | |
876fbba1 MS |
1316 | else |
1317 | DMWARN("%s: unable to establish an integrity profile", | |
1318 | dm_device_name(t->md)); | |
9c47008d MP |
1319 | } |
1320 | ||
ed8b752b MS |
1321 | static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, |
1322 | sector_t start, sector_t len, void *data) | |
1323 | { | |
1324 | unsigned flush = (*(unsigned *)data); | |
1325 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1326 | ||
1327 | return q && (q->flush_flags & flush); | |
1328 | } | |
1329 | ||
1330 | static bool dm_table_supports_flush(struct dm_table *t, unsigned flush) | |
1331 | { | |
1332 | struct dm_target *ti; | |
1333 | unsigned i = 0; | |
1334 | ||
1335 | /* | |
1336 | * Require at least one underlying device to support flushes. | |
1337 | * t->devices includes internal dm devices such as mirror logs | |
1338 | * so we need to use iterate_devices here, which targets | |
1339 | * supporting flushes must provide. | |
1340 | */ | |
1341 | while (i < dm_table_get_num_targets(t)) { | |
1342 | ti = dm_table_get_target(t, i++); | |
1343 | ||
55a62eef | 1344 | if (!ti->num_flush_bios) |
ed8b752b MS |
1345 | continue; |
1346 | ||
0e9c24ed JT |
1347 | if (ti->flush_supported) |
1348 | return 1; | |
1349 | ||
ed8b752b MS |
1350 | if (ti->type->iterate_devices && |
1351 | ti->type->iterate_devices(ti, device_flush_capable, &flush)) | |
1352 | return 1; | |
1353 | } | |
1354 | ||
1355 | return 0; | |
1356 | } | |
1357 | ||
983c7db3 MB |
1358 | static bool dm_table_discard_zeroes_data(struct dm_table *t) |
1359 | { | |
1360 | struct dm_target *ti; | |
1361 | unsigned i = 0; | |
1362 | ||
1363 | /* Ensure that all targets supports discard_zeroes_data. */ | |
1364 | while (i < dm_table_get_num_targets(t)) { | |
1365 | ti = dm_table_get_target(t, i++); | |
1366 | ||
1367 | if (ti->discard_zeroes_data_unsupported) | |
1368 | return 0; | |
1369 | } | |
1370 | ||
1371 | return 1; | |
1372 | } | |
1373 | ||
4693c966 MSB |
1374 | static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev, |
1375 | sector_t start, sector_t len, void *data) | |
1376 | { | |
1377 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1378 | ||
1379 | return q && blk_queue_nonrot(q); | |
1380 | } | |
1381 | ||
c3c4555e MB |
1382 | static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, |
1383 | sector_t start, sector_t len, void *data) | |
1384 | { | |
1385 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1386 | ||
1387 | return q && !blk_queue_add_random(q); | |
1388 | } | |
1389 | ||
1390 | static bool dm_table_all_devices_attribute(struct dm_table *t, | |
1391 | iterate_devices_callout_fn func) | |
4693c966 MSB |
1392 | { |
1393 | struct dm_target *ti; | |
1394 | unsigned i = 0; | |
1395 | ||
4693c966 MSB |
1396 | while (i < dm_table_get_num_targets(t)) { |
1397 | ti = dm_table_get_target(t, i++); | |
1398 | ||
1399 | if (!ti->type->iterate_devices || | |
c3c4555e | 1400 | !ti->type->iterate_devices(ti, func, NULL)) |
4693c966 MSB |
1401 | return 0; |
1402 | } | |
1403 | ||
1404 | return 1; | |
1405 | } | |
1406 | ||
d54eaa5a MS |
1407 | static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev, |
1408 | sector_t start, sector_t len, void *data) | |
1409 | { | |
1410 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1411 | ||
1412 | return q && !q->limits.max_write_same_sectors; | |
1413 | } | |
1414 | ||
1415 | static bool dm_table_supports_write_same(struct dm_table *t) | |
1416 | { | |
1417 | struct dm_target *ti; | |
1418 | unsigned i = 0; | |
1419 | ||
1420 | while (i < dm_table_get_num_targets(t)) { | |
1421 | ti = dm_table_get_target(t, i++); | |
1422 | ||
55a62eef | 1423 | if (!ti->num_write_same_bios) |
d54eaa5a MS |
1424 | return false; |
1425 | ||
1426 | if (!ti->type->iterate_devices || | |
dc019b21 | 1427 | ti->type->iterate_devices(ti, device_not_write_same_capable, NULL)) |
d54eaa5a MS |
1428 | return false; |
1429 | } | |
1430 | ||
1431 | return true; | |
1432 | } | |
1433 | ||
754c5fc7 MS |
1434 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1435 | struct queue_limits *limits) | |
1da177e4 | 1436 | { |
ed8b752b MS |
1437 | unsigned flush = 0; |
1438 | ||
1da177e4 | 1439 | /* |
1197764e | 1440 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1441 | */ |
754c5fc7 | 1442 | q->limits = *limits; |
c9a3f6d6 | 1443 | |
5ae89a87 MS |
1444 | if (!dm_table_supports_discards(t)) |
1445 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1446 | else | |
1447 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1448 | ||
ed8b752b MS |
1449 | if (dm_table_supports_flush(t, REQ_FLUSH)) { |
1450 | flush |= REQ_FLUSH; | |
1451 | if (dm_table_supports_flush(t, REQ_FUA)) | |
1452 | flush |= REQ_FUA; | |
1453 | } | |
1454 | blk_queue_flush(q, flush); | |
1455 | ||
983c7db3 MB |
1456 | if (!dm_table_discard_zeroes_data(t)) |
1457 | q->limits.discard_zeroes_data = 0; | |
1458 | ||
c3c4555e MB |
1459 | /* Ensure that all underlying devices are non-rotational. */ |
1460 | if (dm_table_all_devices_attribute(t, device_is_nonrot)) | |
4693c966 MSB |
1461 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q); |
1462 | else | |
1463 | queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q); | |
1464 | ||
d54eaa5a MS |
1465 | if (!dm_table_supports_write_same(t)) |
1466 | q->limits.max_write_same_sectors = 0; | |
c1a94672 | 1467 | |
9c47008d | 1468 | dm_table_set_integrity(t); |
e6ee8c0b | 1469 | |
c3c4555e MB |
1470 | /* |
1471 | * Determine whether or not this queue's I/O timings contribute | |
1472 | * to the entropy pool, Only request-based targets use this. | |
1473 | * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not | |
1474 | * have it set. | |
1475 | */ | |
1476 | if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random)) | |
1477 | queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, q); | |
1478 | ||
e6ee8c0b KU |
1479 | /* |
1480 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1481 | * visible to other CPUs because, once the flag is set, incoming bios | |
1482 | * are processed by request-based dm, which refers to the queue | |
1483 | * settings. | |
1484 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1485 | * md->deferred where queue settings are not needed yet. | |
1486 | * Those bios are passed to request-based dm at the resume time. | |
1487 | */ | |
1488 | smp_mb(); | |
1489 | if (dm_table_request_based(t)) | |
1490 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1491 | } |
1492 | ||
1493 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1494 | { | |
1495 | return t->num_targets; | |
1496 | } | |
1497 | ||
1498 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1499 | { | |
1500 | return &t->devices; | |
1501 | } | |
1502 | ||
aeb5d727 | 1503 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1504 | { |
1505 | return t->mode; | |
1506 | } | |
08649012 | 1507 | EXPORT_SYMBOL(dm_table_get_mode); |
1da177e4 LT |
1508 | |
1509 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1510 | { | |
1511 | int i = t->num_targets; | |
1512 | struct dm_target *ti = t->targets; | |
1513 | ||
1514 | while (i--) { | |
1515 | if (postsuspend) { | |
1516 | if (ti->type->postsuspend) | |
1517 | ti->type->postsuspend(ti); | |
1518 | } else if (ti->type->presuspend) | |
1519 | ti->type->presuspend(ti); | |
1520 | ||
1521 | ti++; | |
1522 | } | |
1523 | } | |
1524 | ||
1525 | void dm_table_presuspend_targets(struct dm_table *t) | |
1526 | { | |
cf222b37 AK |
1527 | if (!t) |
1528 | return; | |
1529 | ||
e8488d08 | 1530 | suspend_targets(t, 0); |
1da177e4 LT |
1531 | } |
1532 | ||
1533 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1534 | { | |
cf222b37 AK |
1535 | if (!t) |
1536 | return; | |
1537 | ||
e8488d08 | 1538 | suspend_targets(t, 1); |
1da177e4 LT |
1539 | } |
1540 | ||
8757b776 | 1541 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1542 | { |
8757b776 MB |
1543 | int i, r = 0; |
1544 | ||
1545 | for (i = 0; i < t->num_targets; i++) { | |
1546 | struct dm_target *ti = t->targets + i; | |
1547 | ||
1548 | if (!ti->type->preresume) | |
1549 | continue; | |
1550 | ||
1551 | r = ti->type->preresume(ti); | |
7833b08e MS |
1552 | if (r) { |
1553 | DMERR("%s: %s: preresume failed, error = %d", | |
1554 | dm_device_name(t->md), ti->type->name, r); | |
8757b776 | 1555 | return r; |
7833b08e | 1556 | } |
8757b776 | 1557 | } |
1da177e4 LT |
1558 | |
1559 | for (i = 0; i < t->num_targets; i++) { | |
1560 | struct dm_target *ti = t->targets + i; | |
1561 | ||
1562 | if (ti->type->resume) | |
1563 | ti->type->resume(ti); | |
1564 | } | |
8757b776 MB |
1565 | |
1566 | return 0; | |
1da177e4 LT |
1567 | } |
1568 | ||
9d357b07 N |
1569 | void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb) |
1570 | { | |
1571 | list_add(&cb->list, &t->target_callbacks); | |
1572 | } | |
1573 | EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks); | |
1574 | ||
1da177e4 LT |
1575 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) |
1576 | { | |
82b1519b | 1577 | struct dm_dev_internal *dd; |
afb24528 | 1578 | struct list_head *devices = dm_table_get_devices(t); |
9d357b07 | 1579 | struct dm_target_callbacks *cb; |
1da177e4 LT |
1580 | int r = 0; |
1581 | ||
afb24528 | 1582 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1583 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1584 | char b[BDEVNAME_SIZE]; |
1585 | ||
1586 | if (likely(q)) | |
1587 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1588 | else | |
1589 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1590 | dm_device_name(t->md), | |
1591 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1592 | } |
1593 | ||
9d357b07 N |
1594 | list_for_each_entry(cb, &t->target_callbacks, list) |
1595 | if (cb->congested_fn) | |
1596 | r |= cb->congested_fn(cb, bdi_bits); | |
1597 | ||
1da177e4 LT |
1598 | return r; |
1599 | } | |
1600 | ||
cec47e3d KU |
1601 | int dm_table_any_busy_target(struct dm_table *t) |
1602 | { | |
1603 | unsigned i; | |
1604 | struct dm_target *ti; | |
1605 | ||
1606 | for (i = 0; i < t->num_targets; i++) { | |
1607 | ti = t->targets + i; | |
1608 | if (ti->type->busy && ti->type->busy(ti)) | |
1609 | return 1; | |
1610 | } | |
1611 | ||
1612 | return 0; | |
1613 | } | |
1614 | ||
1134e5ae MA |
1615 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1616 | { | |
1134e5ae MA |
1617 | return t->md; |
1618 | } | |
08649012 | 1619 | EXPORT_SYMBOL(dm_table_get_md); |
1134e5ae | 1620 | |
5ae89a87 MS |
1621 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1622 | sector_t start, sector_t len, void *data) | |
1623 | { | |
1624 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1625 | ||
1626 | return q && blk_queue_discard(q); | |
1627 | } | |
1628 | ||
1629 | bool dm_table_supports_discards(struct dm_table *t) | |
1630 | { | |
1631 | struct dm_target *ti; | |
1632 | unsigned i = 0; | |
1633 | ||
5ae89a87 | 1634 | /* |
4c259327 MS |
1635 | * Unless any target used by the table set discards_supported, |
1636 | * require at least one underlying device to support discards. | |
5ae89a87 MS |
1637 | * t->devices includes internal dm devices such as mirror logs |
1638 | * so we need to use iterate_devices here, which targets | |
936688d7 | 1639 | * supporting discard selectively must provide. |
5ae89a87 MS |
1640 | */ |
1641 | while (i < dm_table_get_num_targets(t)) { | |
1642 | ti = dm_table_get_target(t, i++); | |
1643 | ||
55a62eef | 1644 | if (!ti->num_discard_bios) |
936688d7 MS |
1645 | continue; |
1646 | ||
4c259327 MS |
1647 | if (ti->discards_supported) |
1648 | return 1; | |
1649 | ||
5ae89a87 MS |
1650 | if (ti->type->iterate_devices && |
1651 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1652 | return 1; | |
1653 | } | |
1654 | ||
1655 | return 0; | |
1656 | } |