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> |
1da177e4 LT |
20 | #include <asm/atomic.h> |
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 | ||
d5816876 MP |
29 | /* |
30 | * The table has always exactly one reference from either mapped_device->map | |
31 | * or hash_cell->new_map. This reference is not counted in table->holders. | |
32 | * A pair of dm_create_table/dm_destroy_table functions is used for table | |
33 | * creation/destruction. | |
34 | * | |
35 | * Temporary references from the other code increase table->holders. A pair | |
36 | * of dm_table_get/dm_table_put functions is used to manipulate it. | |
37 | * | |
38 | * When the table is about to be destroyed, we wait for table->holders to | |
39 | * drop to zero. | |
40 | */ | |
41 | ||
1da177e4 | 42 | struct dm_table { |
1134e5ae | 43 | struct mapped_device *md; |
1da177e4 | 44 | atomic_t holders; |
e6ee8c0b | 45 | unsigned type; |
1da177e4 LT |
46 | |
47 | /* btree table */ | |
48 | unsigned int depth; | |
49 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | |
50 | sector_t *index[MAX_DEPTH]; | |
51 | ||
52 | unsigned int num_targets; | |
53 | unsigned int num_allocated; | |
54 | sector_t *highs; | |
55 | struct dm_target *targets; | |
56 | ||
5ae89a87 MS |
57 | unsigned discards_supported:1; |
58 | ||
1da177e4 LT |
59 | /* |
60 | * Indicates the rw permissions for the new logical | |
61 | * device. This should be a combination of FMODE_READ | |
62 | * and FMODE_WRITE. | |
63 | */ | |
aeb5d727 | 64 | fmode_t mode; |
1da177e4 LT |
65 | |
66 | /* a list of devices used by this table */ | |
67 | struct list_head devices; | |
68 | ||
1da177e4 LT |
69 | /* events get handed up using this callback */ |
70 | void (*event_fn)(void *); | |
71 | void *event_context; | |
e6ee8c0b KU |
72 | |
73 | struct dm_md_mempools *mempools; | |
1da177e4 LT |
74 | }; |
75 | ||
76 | /* | |
77 | * Similar to ceiling(log_size(n)) | |
78 | */ | |
79 | static unsigned int int_log(unsigned int n, unsigned int base) | |
80 | { | |
81 | int result = 0; | |
82 | ||
83 | while (n > 1) { | |
84 | n = dm_div_up(n, base); | |
85 | result++; | |
86 | } | |
87 | ||
88 | return result; | |
89 | } | |
90 | ||
1da177e4 LT |
91 | /* |
92 | * Calculate the index of the child node of the n'th node k'th key. | |
93 | */ | |
94 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
95 | { | |
96 | return (n * CHILDREN_PER_NODE) + k; | |
97 | } | |
98 | ||
99 | /* | |
100 | * Return the n'th node of level l from table t. | |
101 | */ | |
102 | static inline sector_t *get_node(struct dm_table *t, | |
103 | unsigned int l, unsigned int n) | |
104 | { | |
105 | return t->index[l] + (n * KEYS_PER_NODE); | |
106 | } | |
107 | ||
108 | /* | |
109 | * Return the highest key that you could lookup from the n'th | |
110 | * node on level l of the btree. | |
111 | */ | |
112 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
113 | { | |
114 | for (; l < t->depth - 1; l++) | |
115 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
116 | ||
117 | if (n >= t->counts[l]) | |
118 | return (sector_t) - 1; | |
119 | ||
120 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
121 | } | |
122 | ||
123 | /* | |
124 | * Fills in a level of the btree based on the highs of the level | |
125 | * below it. | |
126 | */ | |
127 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
128 | { | |
129 | unsigned int n, k; | |
130 | sector_t *node; | |
131 | ||
132 | for (n = 0U; n < t->counts[l]; n++) { | |
133 | node = get_node(t, l, n); | |
134 | ||
135 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
136 | node[k] = high(t, l + 1, get_child(n, k)); | |
137 | } | |
138 | ||
139 | return 0; | |
140 | } | |
141 | ||
142 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
143 | { | |
144 | unsigned long size; | |
145 | void *addr; | |
146 | ||
147 | /* | |
148 | * Check that we're not going to overflow. | |
149 | */ | |
150 | if (nmemb > (ULONG_MAX / elem_size)) | |
151 | return NULL; | |
152 | ||
153 | size = nmemb * elem_size; | |
154 | addr = vmalloc(size); | |
155 | if (addr) | |
156 | memset(addr, 0, size); | |
157 | ||
158 | return addr; | |
159 | } | |
160 | ||
161 | /* | |
162 | * highs, and targets are managed as dynamic arrays during a | |
163 | * table load. | |
164 | */ | |
165 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
166 | { | |
167 | sector_t *n_highs; | |
168 | struct dm_target *n_targets; | |
169 | int n = t->num_targets; | |
170 | ||
171 | /* | |
172 | * Allocate both the target array and offset array at once. | |
512875bd JN |
173 | * Append an empty entry to catch sectors beyond the end of |
174 | * the device. | |
1da177e4 | 175 | */ |
512875bd | 176 | n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) + |
1da177e4 LT |
177 | sizeof(sector_t)); |
178 | if (!n_highs) | |
179 | return -ENOMEM; | |
180 | ||
181 | n_targets = (struct dm_target *) (n_highs + num); | |
182 | ||
183 | if (n) { | |
184 | memcpy(n_highs, t->highs, sizeof(*n_highs) * n); | |
185 | memcpy(n_targets, t->targets, sizeof(*n_targets) * n); | |
186 | } | |
187 | ||
188 | memset(n_highs + n, -1, sizeof(*n_highs) * (num - n)); | |
189 | vfree(t->highs); | |
190 | ||
191 | t->num_allocated = num; | |
192 | t->highs = n_highs; | |
193 | t->targets = n_targets; | |
194 | ||
195 | return 0; | |
196 | } | |
197 | ||
aeb5d727 | 198 | int dm_table_create(struct dm_table **result, fmode_t mode, |
1134e5ae | 199 | unsigned num_targets, struct mapped_device *md) |
1da177e4 | 200 | { |
094262db | 201 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
202 | |
203 | if (!t) | |
204 | return -ENOMEM; | |
205 | ||
1da177e4 | 206 | INIT_LIST_HEAD(&t->devices); |
d5816876 | 207 | atomic_set(&t->holders, 0); |
5ae89a87 | 208 | t->discards_supported = 1; |
1da177e4 LT |
209 | |
210 | if (!num_targets) | |
211 | num_targets = KEYS_PER_NODE; | |
212 | ||
213 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
214 | ||
215 | if (alloc_targets(t, num_targets)) { | |
216 | kfree(t); | |
217 | t = NULL; | |
218 | return -ENOMEM; | |
219 | } | |
220 | ||
221 | t->mode = mode; | |
1134e5ae | 222 | t->md = md; |
1da177e4 LT |
223 | *result = t; |
224 | return 0; | |
225 | } | |
226 | ||
227 | static void free_devices(struct list_head *devices) | |
228 | { | |
229 | struct list_head *tmp, *next; | |
230 | ||
afb24528 | 231 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
232 | struct dm_dev_internal *dd = |
233 | list_entry(tmp, struct dm_dev_internal, list); | |
1b6da754 JB |
234 | DMWARN("dm_table_destroy: dm_put_device call missing for %s", |
235 | dd->dm_dev.name); | |
1da177e4 LT |
236 | kfree(dd); |
237 | } | |
238 | } | |
239 | ||
d5816876 | 240 | void dm_table_destroy(struct dm_table *t) |
1da177e4 LT |
241 | { |
242 | unsigned int i; | |
243 | ||
a7940155 AK |
244 | if (!t) |
245 | return; | |
246 | ||
d5816876 MP |
247 | while (atomic_read(&t->holders)) |
248 | msleep(1); | |
249 | smp_mb(); | |
250 | ||
26803b9f | 251 | /* free the indexes */ |
1da177e4 LT |
252 | if (t->depth >= 2) |
253 | vfree(t->index[t->depth - 2]); | |
254 | ||
255 | /* free the targets */ | |
256 | for (i = 0; i < t->num_targets; i++) { | |
257 | struct dm_target *tgt = t->targets + i; | |
258 | ||
259 | if (tgt->type->dtr) | |
260 | tgt->type->dtr(tgt); | |
261 | ||
262 | dm_put_target_type(tgt->type); | |
263 | } | |
264 | ||
265 | vfree(t->highs); | |
266 | ||
267 | /* free the device list */ | |
1b6da754 | 268 | if (t->devices.next != &t->devices) |
1da177e4 | 269 | free_devices(&t->devices); |
1da177e4 | 270 | |
e6ee8c0b KU |
271 | dm_free_md_mempools(t->mempools); |
272 | ||
1da177e4 LT |
273 | kfree(t); |
274 | } | |
275 | ||
276 | void dm_table_get(struct dm_table *t) | |
277 | { | |
278 | atomic_inc(&t->holders); | |
279 | } | |
280 | ||
281 | void dm_table_put(struct dm_table *t) | |
282 | { | |
283 | if (!t) | |
284 | return; | |
285 | ||
d5816876 MP |
286 | smp_mb__before_atomic_dec(); |
287 | atomic_dec(&t->holders); | |
1da177e4 LT |
288 | } |
289 | ||
290 | /* | |
291 | * Checks to see if we need to extend highs or targets. | |
292 | */ | |
293 | static inline int check_space(struct dm_table *t) | |
294 | { | |
295 | if (t->num_targets >= t->num_allocated) | |
296 | return alloc_targets(t, t->num_allocated * 2); | |
297 | ||
298 | return 0; | |
299 | } | |
300 | ||
1da177e4 LT |
301 | /* |
302 | * See if we've already got a device in the list. | |
303 | */ | |
82b1519b | 304 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 305 | { |
82b1519b | 306 | struct dm_dev_internal *dd; |
1da177e4 LT |
307 | |
308 | list_for_each_entry (dd, l, list) | |
82b1519b | 309 | if (dd->dm_dev.bdev->bd_dev == dev) |
1da177e4 LT |
310 | return dd; |
311 | ||
312 | return NULL; | |
313 | } | |
314 | ||
315 | /* | |
316 | * Open a device so we can use it as a map destination. | |
317 | */ | |
82b1519b MP |
318 | static int open_dev(struct dm_dev_internal *d, dev_t dev, |
319 | struct mapped_device *md) | |
1da177e4 LT |
320 | { |
321 | static char *_claim_ptr = "I belong to device-mapper"; | |
322 | struct block_device *bdev; | |
323 | ||
324 | int r; | |
325 | ||
82b1519b | 326 | BUG_ON(d->dm_dev.bdev); |
1da177e4 | 327 | |
82b1519b | 328 | bdev = open_by_devnum(dev, d->dm_dev.mode); |
1da177e4 LT |
329 | if (IS_ERR(bdev)) |
330 | return PTR_ERR(bdev); | |
f165921d | 331 | r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md)); |
1da177e4 | 332 | if (r) |
9a1c3542 | 333 | blkdev_put(bdev, d->dm_dev.mode); |
1da177e4 | 334 | else |
82b1519b | 335 | d->dm_dev.bdev = bdev; |
1da177e4 LT |
336 | return r; |
337 | } | |
338 | ||
339 | /* | |
340 | * Close a device that we've been using. | |
341 | */ | |
82b1519b | 342 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 343 | { |
82b1519b | 344 | if (!d->dm_dev.bdev) |
1da177e4 LT |
345 | return; |
346 | ||
82b1519b | 347 | bd_release_from_disk(d->dm_dev.bdev, dm_disk(md)); |
9a1c3542 | 348 | blkdev_put(d->dm_dev.bdev, d->dm_dev.mode); |
82b1519b | 349 | d->dm_dev.bdev = NULL; |
1da177e4 LT |
350 | } |
351 | ||
352 | /* | |
f6a1ed10 | 353 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 354 | */ |
f6a1ed10 MP |
355 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
356 | sector_t start, sector_t len, void *data) | |
1da177e4 | 357 | { |
754c5fc7 MS |
358 | struct queue_limits *limits = data; |
359 | struct block_device *bdev = dev->bdev; | |
360 | sector_t dev_size = | |
361 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 362 | unsigned short logical_block_size_sectors = |
754c5fc7 | 363 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 364 | char b[BDEVNAME_SIZE]; |
2cd54d9b MA |
365 | |
366 | if (!dev_size) | |
f6a1ed10 | 367 | return 0; |
2cd54d9b | 368 | |
5dea271b | 369 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
370 | DMWARN("%s: %s too small for target: " |
371 | "start=%llu, len=%llu, dev_size=%llu", | |
372 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
373 | (unsigned long long)start, | |
374 | (unsigned long long)len, | |
375 | (unsigned long long)dev_size); | |
f6a1ed10 | 376 | return 1; |
02acc3a4 MS |
377 | } |
378 | ||
379 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 380 | return 0; |
02acc3a4 MS |
381 | |
382 | if (start & (logical_block_size_sectors - 1)) { | |
383 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 384 | "logical block size %u of %s", |
02acc3a4 MS |
385 | dm_device_name(ti->table->md), |
386 | (unsigned long long)start, | |
754c5fc7 | 387 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 388 | return 1; |
02acc3a4 MS |
389 | } |
390 | ||
5dea271b | 391 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 392 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 393 | "logical block size %u of %s", |
02acc3a4 | 394 | dm_device_name(ti->table->md), |
5dea271b | 395 | (unsigned long long)len, |
754c5fc7 | 396 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 397 | return 1; |
02acc3a4 MS |
398 | } |
399 | ||
f6a1ed10 | 400 | return 0; |
1da177e4 LT |
401 | } |
402 | ||
403 | /* | |
570b9d96 | 404 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 405 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
406 | * device and not to touch the existing bdev field in case |
407 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 408 | */ |
aeb5d727 | 409 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 410 | struct mapped_device *md) |
1da177e4 LT |
411 | { |
412 | int r; | |
570b9d96 | 413 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 414 | |
570b9d96 AK |
415 | dd_new = dd_old = *dd; |
416 | ||
417 | dd_new.dm_dev.mode |= new_mode; | |
418 | dd_new.dm_dev.bdev = NULL; | |
419 | ||
420 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
421 | if (r) | |
422 | return r; | |
1da177e4 | 423 | |
82b1519b | 424 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 425 | close_dev(&dd_old, md); |
1da177e4 | 426 | |
570b9d96 | 427 | return 0; |
1da177e4 LT |
428 | } |
429 | ||
430 | /* | |
431 | * Add a device to the list, or just increment the usage count if | |
432 | * it's already present. | |
433 | */ | |
434 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
8215d6ec | 435 | const char *path, fmode_t mode, struct dm_dev **result) |
1da177e4 LT |
436 | { |
437 | int r; | |
69a2ce72 | 438 | dev_t uninitialized_var(dev); |
82b1519b | 439 | struct dm_dev_internal *dd; |
1da177e4 LT |
440 | unsigned int major, minor; |
441 | ||
547bc926 | 442 | BUG_ON(!t); |
1da177e4 LT |
443 | |
444 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
445 | /* Extract the major/minor numbers */ | |
446 | dev = MKDEV(major, minor); | |
447 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
448 | return -EOVERFLOW; | |
449 | } else { | |
450 | /* convert the path to a device */ | |
72e8264e CH |
451 | struct block_device *bdev = lookup_bdev(path); |
452 | ||
453 | if (IS_ERR(bdev)) | |
454 | return PTR_ERR(bdev); | |
455 | dev = bdev->bd_dev; | |
456 | bdput(bdev); | |
1da177e4 LT |
457 | } |
458 | ||
459 | dd = find_device(&t->devices, dev); | |
460 | if (!dd) { | |
461 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
462 | if (!dd) | |
463 | return -ENOMEM; | |
464 | ||
82b1519b MP |
465 | dd->dm_dev.mode = mode; |
466 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 467 | |
f165921d | 468 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
469 | kfree(dd); |
470 | return r; | |
471 | } | |
472 | ||
82b1519b | 473 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
474 | |
475 | atomic_set(&dd->count, 0); | |
476 | list_add(&dd->list, &t->devices); | |
477 | ||
82b1519b | 478 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 479 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
480 | if (r) |
481 | return r; | |
482 | } | |
483 | atomic_inc(&dd->count); | |
484 | ||
82b1519b | 485 | *result = &dd->dm_dev; |
1da177e4 LT |
486 | return 0; |
487 | } | |
488 | ||
754c5fc7 | 489 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 490 | sector_t start, sector_t len, void *data) |
1da177e4 | 491 | { |
754c5fc7 MS |
492 | struct queue_limits *limits = data; |
493 | struct block_device *bdev = dev->bdev; | |
165125e1 | 494 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
495 | char b[BDEVNAME_SIZE]; |
496 | ||
497 | if (unlikely(!q)) { | |
498 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
499 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 500 | return 0; |
0c2322e4 | 501 | } |
3cb40214 | 502 | |
b27d7f16 MP |
503 | if (bdev_stack_limits(limits, bdev, start) < 0) |
504 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
505 | "physical_block_size=%u, logical_block_size=%u, " |
506 | "alignment_offset=%u, start=%llu", | |
507 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
508 | q->limits.physical_block_size, | |
509 | q->limits.logical_block_size, | |
510 | q->limits.alignment_offset, | |
b27d7f16 | 511 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 512 | |
9980c638 MB |
513 | /* |
514 | * Check if merge fn is supported. | |
515 | * If not we'll force DM to use PAGE_SIZE or | |
516 | * smaller I/O, just to be safe. | |
3cb40214 | 517 | */ |
9980c638 MB |
518 | |
519 | if (q->merge_bvec_fn && !ti->type->merge) | |
754c5fc7 MS |
520 | limits->max_sectors = |
521 | min_not_zero(limits->max_sectors, | |
3cb40214 | 522 | (unsigned int) (PAGE_SIZE >> 9)); |
754c5fc7 | 523 | return 0; |
3cb40214 BR |
524 | } |
525 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 526 | |
8215d6ec NK |
527 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
528 | struct dm_dev **result) | |
3cb40214 | 529 | { |
8215d6ec | 530 | return __table_get_device(ti->table, ti, path, mode, result); |
1da177e4 LT |
531 | } |
532 | ||
754c5fc7 | 533 | |
1da177e4 LT |
534 | /* |
535 | * Decrement a devices use count and remove it if necessary. | |
536 | */ | |
82b1519b | 537 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 538 | { |
82b1519b MP |
539 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
540 | dm_dev); | |
541 | ||
1da177e4 | 542 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 543 | close_dev(dd, ti->table->md); |
1da177e4 LT |
544 | list_del(&dd->list); |
545 | kfree(dd); | |
546 | } | |
547 | } | |
548 | ||
549 | /* | |
550 | * Checks to see if the target joins onto the end of the table. | |
551 | */ | |
552 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
553 | { | |
554 | struct dm_target *prev; | |
555 | ||
556 | if (!table->num_targets) | |
557 | return !ti->begin; | |
558 | ||
559 | prev = &table->targets[table->num_targets - 1]; | |
560 | return (ti->begin == (prev->begin + prev->len)); | |
561 | } | |
562 | ||
563 | /* | |
564 | * Used to dynamically allocate the arg array. | |
565 | */ | |
566 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
567 | { | |
568 | char **argv; | |
569 | unsigned new_size; | |
570 | ||
571 | new_size = *array_size ? *array_size * 2 : 64; | |
572 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
573 | if (argv) { | |
574 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
575 | *array_size = new_size; | |
576 | } | |
577 | ||
578 | kfree(old_argv); | |
579 | return argv; | |
580 | } | |
581 | ||
582 | /* | |
583 | * Destructively splits up the argument list to pass to ctr. | |
584 | */ | |
585 | int dm_split_args(int *argc, char ***argvp, char *input) | |
586 | { | |
587 | char *start, *end = input, *out, **argv = NULL; | |
588 | unsigned array_size = 0; | |
589 | ||
590 | *argc = 0; | |
814d6862 DT |
591 | |
592 | if (!input) { | |
593 | *argvp = NULL; | |
594 | return 0; | |
595 | } | |
596 | ||
1da177e4 LT |
597 | argv = realloc_argv(&array_size, argv); |
598 | if (!argv) | |
599 | return -ENOMEM; | |
600 | ||
601 | while (1) { | |
1da177e4 | 602 | /* Skip whitespace */ |
e7d2860b | 603 | start = skip_spaces(end); |
1da177e4 LT |
604 | |
605 | if (!*start) | |
606 | break; /* success, we hit the end */ | |
607 | ||
608 | /* 'out' is used to remove any back-quotes */ | |
609 | end = out = start; | |
610 | while (*end) { | |
611 | /* Everything apart from '\0' can be quoted */ | |
612 | if (*end == '\\' && *(end + 1)) { | |
613 | *out++ = *(end + 1); | |
614 | end += 2; | |
615 | continue; | |
616 | } | |
617 | ||
618 | if (isspace(*end)) | |
619 | break; /* end of token */ | |
620 | ||
621 | *out++ = *end++; | |
622 | } | |
623 | ||
624 | /* have we already filled the array ? */ | |
625 | if ((*argc + 1) > array_size) { | |
626 | argv = realloc_argv(&array_size, argv); | |
627 | if (!argv) | |
628 | return -ENOMEM; | |
629 | } | |
630 | ||
631 | /* we know this is whitespace */ | |
632 | if (*end) | |
633 | end++; | |
634 | ||
635 | /* terminate the string and put it in the array */ | |
636 | *out = '\0'; | |
637 | argv[*argc] = start; | |
638 | (*argc)++; | |
639 | } | |
640 | ||
641 | *argvp = argv; | |
642 | return 0; | |
643 | } | |
644 | ||
be6d4305 MS |
645 | /* |
646 | * Impose necessary and sufficient conditions on a devices's table such | |
647 | * that any incoming bio which respects its logical_block_size can be | |
648 | * processed successfully. If it falls across the boundary between | |
649 | * two or more targets, the size of each piece it gets split into must | |
650 | * be compatible with the logical_block_size of the target processing it. | |
651 | */ | |
754c5fc7 MS |
652 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
653 | struct queue_limits *limits) | |
be6d4305 MS |
654 | { |
655 | /* | |
656 | * This function uses arithmetic modulo the logical_block_size | |
657 | * (in units of 512-byte sectors). | |
658 | */ | |
659 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 660 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
661 | |
662 | /* | |
663 | * Offset of the start of the next table entry, mod logical_block_size. | |
664 | */ | |
665 | unsigned short next_target_start = 0; | |
666 | ||
667 | /* | |
668 | * Given an aligned bio that extends beyond the end of a | |
669 | * target, how many sectors must the next target handle? | |
670 | */ | |
671 | unsigned short remaining = 0; | |
672 | ||
673 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 674 | struct queue_limits ti_limits; |
be6d4305 MS |
675 | unsigned i = 0; |
676 | ||
677 | /* | |
678 | * Check each entry in the table in turn. | |
679 | */ | |
680 | while (i < dm_table_get_num_targets(table)) { | |
681 | ti = dm_table_get_target(table, i++); | |
682 | ||
754c5fc7 MS |
683 | blk_set_default_limits(&ti_limits); |
684 | ||
685 | /* combine all target devices' limits */ | |
686 | if (ti->type->iterate_devices) | |
687 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
688 | &ti_limits); | |
689 | ||
be6d4305 MS |
690 | /* |
691 | * If the remaining sectors fall entirely within this | |
692 | * table entry are they compatible with its logical_block_size? | |
693 | */ | |
694 | if (remaining < ti->len && | |
754c5fc7 | 695 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
696 | SECTOR_SHIFT) - 1)) |
697 | break; /* Error */ | |
698 | ||
699 | next_target_start = | |
700 | (unsigned short) ((next_target_start + ti->len) & | |
701 | (device_logical_block_size_sects - 1)); | |
702 | remaining = next_target_start ? | |
703 | device_logical_block_size_sects - next_target_start : 0; | |
704 | } | |
705 | ||
706 | if (remaining) { | |
707 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 708 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
709 | dm_device_name(table->md), i, |
710 | (unsigned long long) ti->begin, | |
711 | (unsigned long long) ti->len, | |
754c5fc7 | 712 | limits->logical_block_size); |
be6d4305 MS |
713 | return -EINVAL; |
714 | } | |
715 | ||
716 | return 0; | |
717 | } | |
718 | ||
1da177e4 LT |
719 | int dm_table_add_target(struct dm_table *t, const char *type, |
720 | sector_t start, sector_t len, char *params) | |
721 | { | |
722 | int r = -EINVAL, argc; | |
723 | char **argv; | |
724 | struct dm_target *tgt; | |
725 | ||
726 | if ((r = check_space(t))) | |
727 | return r; | |
728 | ||
729 | tgt = t->targets + t->num_targets; | |
730 | memset(tgt, 0, sizeof(*tgt)); | |
731 | ||
732 | if (!len) { | |
72d94861 | 733 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
734 | return -EINVAL; |
735 | } | |
736 | ||
737 | tgt->type = dm_get_target_type(type); | |
738 | if (!tgt->type) { | |
72d94861 AK |
739 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
740 | type); | |
1da177e4 LT |
741 | return -EINVAL; |
742 | } | |
743 | ||
744 | tgt->table = t; | |
745 | tgt->begin = start; | |
746 | tgt->len = len; | |
747 | tgt->error = "Unknown error"; | |
748 | ||
749 | /* | |
750 | * Does this target adjoin the previous one ? | |
751 | */ | |
752 | if (!adjoin(t, tgt)) { | |
753 | tgt->error = "Gap in table"; | |
754 | r = -EINVAL; | |
755 | goto bad; | |
756 | } | |
757 | ||
758 | r = dm_split_args(&argc, &argv, params); | |
759 | if (r) { | |
760 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
761 | goto bad; | |
762 | } | |
763 | ||
764 | r = tgt->type->ctr(tgt, argc, argv); | |
765 | kfree(argv); | |
766 | if (r) | |
767 | goto bad; | |
768 | ||
769 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
770 | ||
5ae89a87 MS |
771 | if (!tgt->num_discard_requests) |
772 | t->discards_supported = 0; | |
773 | ||
1da177e4 LT |
774 | return 0; |
775 | ||
776 | bad: | |
72d94861 | 777 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
778 | dm_put_target_type(tgt->type); |
779 | return r; | |
780 | } | |
781 | ||
26803b9f | 782 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
783 | { |
784 | unsigned i; | |
785 | unsigned bio_based = 0, request_based = 0; | |
786 | struct dm_target *tgt; | |
787 | struct dm_dev_internal *dd; | |
788 | struct list_head *devices; | |
789 | ||
790 | for (i = 0; i < t->num_targets; i++) { | |
791 | tgt = t->targets + i; | |
792 | if (dm_target_request_based(tgt)) | |
793 | request_based = 1; | |
794 | else | |
795 | bio_based = 1; | |
796 | ||
797 | if (bio_based && request_based) { | |
798 | DMWARN("Inconsistent table: different target types" | |
799 | " can't be mixed up"); | |
800 | return -EINVAL; | |
801 | } | |
802 | } | |
803 | ||
804 | if (bio_based) { | |
805 | /* We must use this table as bio-based */ | |
806 | t->type = DM_TYPE_BIO_BASED; | |
807 | return 0; | |
808 | } | |
809 | ||
810 | BUG_ON(!request_based); /* No targets in this table */ | |
811 | ||
812 | /* Non-request-stackable devices can't be used for request-based dm */ | |
813 | devices = dm_table_get_devices(t); | |
814 | list_for_each_entry(dd, devices, list) { | |
815 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
816 | DMWARN("table load rejected: including" | |
817 | " non-request-stackable devices"); | |
818 | return -EINVAL; | |
819 | } | |
820 | } | |
821 | ||
822 | /* | |
823 | * Request-based dm supports only tables that have a single target now. | |
824 | * To support multiple targets, request splitting support is needed, | |
825 | * and that needs lots of changes in the block-layer. | |
826 | * (e.g. request completion process for partial completion.) | |
827 | */ | |
828 | if (t->num_targets > 1) { | |
829 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
830 | return -EINVAL; | |
831 | } | |
832 | ||
833 | t->type = DM_TYPE_REQUEST_BASED; | |
834 | ||
835 | return 0; | |
836 | } | |
837 | ||
838 | unsigned dm_table_get_type(struct dm_table *t) | |
839 | { | |
840 | return t->type; | |
841 | } | |
842 | ||
843 | bool dm_table_request_based(struct dm_table *t) | |
844 | { | |
845 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
846 | } | |
847 | ||
848 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
849 | { | |
850 | unsigned type = dm_table_get_type(t); | |
851 | ||
852 | if (unlikely(type == DM_TYPE_NONE)) { | |
853 | DMWARN("no table type is set, can't allocate mempools"); | |
854 | return -EINVAL; | |
855 | } | |
856 | ||
857 | t->mempools = dm_alloc_md_mempools(type); | |
858 | if (!t->mempools) | |
859 | return -ENOMEM; | |
860 | ||
861 | return 0; | |
862 | } | |
863 | ||
864 | void dm_table_free_md_mempools(struct dm_table *t) | |
865 | { | |
866 | dm_free_md_mempools(t->mempools); | |
867 | t->mempools = NULL; | |
868 | } | |
869 | ||
870 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
871 | { | |
872 | return t->mempools; | |
873 | } | |
874 | ||
1da177e4 LT |
875 | static int setup_indexes(struct dm_table *t) |
876 | { | |
877 | int i; | |
878 | unsigned int total = 0; | |
879 | sector_t *indexes; | |
880 | ||
881 | /* allocate the space for *all* the indexes */ | |
882 | for (i = t->depth - 2; i >= 0; i--) { | |
883 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
884 | total += t->counts[i]; | |
885 | } | |
886 | ||
887 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
888 | if (!indexes) | |
889 | return -ENOMEM; | |
890 | ||
891 | /* set up internal nodes, bottom-up */ | |
82d601dc | 892 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
893 | t->index[i] = indexes; |
894 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
895 | setup_btree_index(i, t); | |
896 | } | |
897 | ||
898 | return 0; | |
899 | } | |
900 | ||
901 | /* | |
902 | * Builds the btree to index the map. | |
903 | */ | |
26803b9f | 904 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
905 | { |
906 | int r = 0; | |
907 | unsigned int leaf_nodes; | |
908 | ||
1da177e4 LT |
909 | /* how many indexes will the btree have ? */ |
910 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
911 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
912 | ||
913 | /* leaf layer has already been set up */ | |
914 | t->counts[t->depth - 1] = leaf_nodes; | |
915 | t->index[t->depth - 1] = t->highs; | |
916 | ||
917 | if (t->depth >= 2) | |
918 | r = setup_indexes(t); | |
919 | ||
920 | return r; | |
921 | } | |
922 | ||
26803b9f WD |
923 | /* |
924 | * Register the mapped device for blk_integrity support if | |
925 | * the underlying devices support it. | |
926 | */ | |
927 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
928 | { | |
929 | struct list_head *devices = dm_table_get_devices(t); | |
930 | struct dm_dev_internal *dd; | |
931 | ||
932 | list_for_each_entry(dd, devices, list) | |
933 | if (bdev_get_integrity(dd->dm_dev.bdev)) | |
934 | return blk_integrity_register(dm_disk(md), NULL); | |
935 | ||
936 | return 0; | |
937 | } | |
938 | ||
939 | /* | |
940 | * Prepares the table for use by building the indices, | |
941 | * setting the type, and allocating mempools. | |
942 | */ | |
943 | int dm_table_complete(struct dm_table *t) | |
944 | { | |
945 | int r; | |
946 | ||
947 | r = dm_table_set_type(t); | |
948 | if (r) { | |
949 | DMERR("unable to set table type"); | |
950 | return r; | |
951 | } | |
952 | ||
953 | r = dm_table_build_index(t); | |
954 | if (r) { | |
955 | DMERR("unable to build btrees"); | |
956 | return r; | |
957 | } | |
958 | ||
959 | r = dm_table_prealloc_integrity(t, t->md); | |
960 | if (r) { | |
961 | DMERR("could not register integrity profile."); | |
962 | return r; | |
963 | } | |
964 | ||
965 | r = dm_table_alloc_md_mempools(t); | |
966 | if (r) | |
967 | DMERR("unable to allocate mempools"); | |
968 | ||
969 | return r; | |
970 | } | |
971 | ||
48c9c27b | 972 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
973 | void dm_table_event_callback(struct dm_table *t, |
974 | void (*fn)(void *), void *context) | |
975 | { | |
48c9c27b | 976 | mutex_lock(&_event_lock); |
1da177e4 LT |
977 | t->event_fn = fn; |
978 | t->event_context = context; | |
48c9c27b | 979 | mutex_unlock(&_event_lock); |
1da177e4 LT |
980 | } |
981 | ||
982 | void dm_table_event(struct dm_table *t) | |
983 | { | |
984 | /* | |
985 | * You can no longer call dm_table_event() from interrupt | |
986 | * context, use a bottom half instead. | |
987 | */ | |
988 | BUG_ON(in_interrupt()); | |
989 | ||
48c9c27b | 990 | mutex_lock(&_event_lock); |
1da177e4 LT |
991 | if (t->event_fn) |
992 | t->event_fn(t->event_context); | |
48c9c27b | 993 | mutex_unlock(&_event_lock); |
1da177e4 LT |
994 | } |
995 | ||
996 | sector_t dm_table_get_size(struct dm_table *t) | |
997 | { | |
998 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
999 | } | |
1000 | ||
1001 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1002 | { | |
14353539 | 1003 | if (index >= t->num_targets) |
1da177e4 LT |
1004 | return NULL; |
1005 | ||
1006 | return t->targets + index; | |
1007 | } | |
1008 | ||
1009 | /* | |
1010 | * Search the btree for the correct target. | |
512875bd JN |
1011 | * |
1012 | * Caller should check returned pointer with dm_target_is_valid() | |
1013 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1014 | */ |
1015 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1016 | { | |
1017 | unsigned int l, n = 0, k = 0; | |
1018 | sector_t *node; | |
1019 | ||
1020 | for (l = 0; l < t->depth; l++) { | |
1021 | n = get_child(n, k); | |
1022 | node = get_node(t, l, n); | |
1023 | ||
1024 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1025 | if (node[k] >= sector) | |
1026 | break; | |
1027 | } | |
1028 | ||
1029 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1030 | } | |
1031 | ||
754c5fc7 MS |
1032 | /* |
1033 | * Establish the new table's queue_limits and validate them. | |
1034 | */ | |
1035 | int dm_calculate_queue_limits(struct dm_table *table, | |
1036 | struct queue_limits *limits) | |
1037 | { | |
1038 | struct dm_target *uninitialized_var(ti); | |
1039 | struct queue_limits ti_limits; | |
1040 | unsigned i = 0; | |
1041 | ||
1042 | blk_set_default_limits(limits); | |
1043 | ||
1044 | while (i < dm_table_get_num_targets(table)) { | |
1045 | blk_set_default_limits(&ti_limits); | |
1046 | ||
1047 | ti = dm_table_get_target(table, i++); | |
1048 | ||
1049 | if (!ti->type->iterate_devices) | |
1050 | goto combine_limits; | |
1051 | ||
1052 | /* | |
1053 | * Combine queue limits of all the devices this target uses. | |
1054 | */ | |
1055 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1056 | &ti_limits); | |
1057 | ||
40bea431 MS |
1058 | /* Set I/O hints portion of queue limits */ |
1059 | if (ti->type->io_hints) | |
1060 | ti->type->io_hints(ti, &ti_limits); | |
1061 | ||
754c5fc7 MS |
1062 | /* |
1063 | * Check each device area is consistent with the target's | |
1064 | * overall queue limits. | |
1065 | */ | |
f6a1ed10 MP |
1066 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1067 | &ti_limits)) | |
754c5fc7 MS |
1068 | return -EINVAL; |
1069 | ||
1070 | combine_limits: | |
1071 | /* | |
1072 | * Merge this target's queue limits into the overall limits | |
1073 | * for the table. | |
1074 | */ | |
1075 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1076 | DMWARN("%s: adding target device " |
754c5fc7 | 1077 | "(start sect %llu len %llu) " |
b27d7f16 | 1078 | "caused an alignment inconsistency", |
754c5fc7 MS |
1079 | dm_device_name(table->md), |
1080 | (unsigned long long) ti->begin, | |
1081 | (unsigned long long) ti->len); | |
1082 | } | |
1083 | ||
1084 | return validate_hardware_logical_block_alignment(table, limits); | |
1085 | } | |
1086 | ||
9c47008d MP |
1087 | /* |
1088 | * Set the integrity profile for this device if all devices used have | |
1089 | * matching profiles. | |
1090 | */ | |
1091 | static void dm_table_set_integrity(struct dm_table *t) | |
1092 | { | |
1093 | struct list_head *devices = dm_table_get_devices(t); | |
1094 | struct dm_dev_internal *prev = NULL, *dd = NULL; | |
1095 | ||
1096 | if (!blk_get_integrity(dm_disk(t->md))) | |
1097 | return; | |
1098 | ||
1099 | list_for_each_entry(dd, devices, list) { | |
1100 | if (prev && | |
1101 | blk_integrity_compare(prev->dm_dev.bdev->bd_disk, | |
1102 | dd->dm_dev.bdev->bd_disk) < 0) { | |
1103 | DMWARN("%s: integrity not set: %s and %s mismatch", | |
1104 | dm_device_name(t->md), | |
1105 | prev->dm_dev.bdev->bd_disk->disk_name, | |
1106 | dd->dm_dev.bdev->bd_disk->disk_name); | |
1107 | goto no_integrity; | |
1108 | } | |
1109 | prev = dd; | |
1110 | } | |
1111 | ||
1112 | if (!prev || !bdev_get_integrity(prev->dm_dev.bdev)) | |
1113 | goto no_integrity; | |
1114 | ||
1115 | blk_integrity_register(dm_disk(t->md), | |
1116 | bdev_get_integrity(prev->dm_dev.bdev)); | |
1117 | ||
1118 | return; | |
1119 | ||
1120 | no_integrity: | |
1121 | blk_integrity_register(dm_disk(t->md), NULL); | |
1122 | ||
1123 | return; | |
1124 | } | |
1125 | ||
754c5fc7 MS |
1126 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1127 | struct queue_limits *limits) | |
1da177e4 LT |
1128 | { |
1129 | /* | |
1197764e | 1130 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1131 | */ |
754c5fc7 | 1132 | q->limits = *limits; |
c9a3f6d6 | 1133 | |
754c5fc7 | 1134 | if (limits->no_cluster) |
c9a3f6d6 | 1135 | queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 1136 | else |
c9a3f6d6 | 1137 | queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 1138 | |
5ae89a87 MS |
1139 | if (!dm_table_supports_discards(t)) |
1140 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1141 | else | |
1142 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1143 | ||
9c47008d | 1144 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1145 | |
1146 | /* | |
1147 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1148 | * visible to other CPUs because, once the flag is set, incoming bios | |
1149 | * are processed by request-based dm, which refers to the queue | |
1150 | * settings. | |
1151 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1152 | * md->deferred where queue settings are not needed yet. | |
1153 | * Those bios are passed to request-based dm at the resume time. | |
1154 | */ | |
1155 | smp_mb(); | |
1156 | if (dm_table_request_based(t)) | |
1157 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1158 | } |
1159 | ||
1160 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1161 | { | |
1162 | return t->num_targets; | |
1163 | } | |
1164 | ||
1165 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1166 | { | |
1167 | return &t->devices; | |
1168 | } | |
1169 | ||
aeb5d727 | 1170 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1171 | { |
1172 | return t->mode; | |
1173 | } | |
1174 | ||
1175 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1176 | { | |
1177 | int i = t->num_targets; | |
1178 | struct dm_target *ti = t->targets; | |
1179 | ||
1180 | while (i--) { | |
1181 | if (postsuspend) { | |
1182 | if (ti->type->postsuspend) | |
1183 | ti->type->postsuspend(ti); | |
1184 | } else if (ti->type->presuspend) | |
1185 | ti->type->presuspend(ti); | |
1186 | ||
1187 | ti++; | |
1188 | } | |
1189 | } | |
1190 | ||
1191 | void dm_table_presuspend_targets(struct dm_table *t) | |
1192 | { | |
cf222b37 AK |
1193 | if (!t) |
1194 | return; | |
1195 | ||
e8488d08 | 1196 | suspend_targets(t, 0); |
1da177e4 LT |
1197 | } |
1198 | ||
1199 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1200 | { | |
cf222b37 AK |
1201 | if (!t) |
1202 | return; | |
1203 | ||
e8488d08 | 1204 | suspend_targets(t, 1); |
1da177e4 LT |
1205 | } |
1206 | ||
8757b776 | 1207 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1208 | { |
8757b776 MB |
1209 | int i, r = 0; |
1210 | ||
1211 | for (i = 0; i < t->num_targets; i++) { | |
1212 | struct dm_target *ti = t->targets + i; | |
1213 | ||
1214 | if (!ti->type->preresume) | |
1215 | continue; | |
1216 | ||
1217 | r = ti->type->preresume(ti); | |
1218 | if (r) | |
1219 | return r; | |
1220 | } | |
1da177e4 LT |
1221 | |
1222 | for (i = 0; i < t->num_targets; i++) { | |
1223 | struct dm_target *ti = t->targets + i; | |
1224 | ||
1225 | if (ti->type->resume) | |
1226 | ti->type->resume(ti); | |
1227 | } | |
8757b776 MB |
1228 | |
1229 | return 0; | |
1da177e4 LT |
1230 | } |
1231 | ||
1232 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | |
1233 | { | |
82b1519b | 1234 | struct dm_dev_internal *dd; |
afb24528 | 1235 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 LT |
1236 | int r = 0; |
1237 | ||
afb24528 | 1238 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1239 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1240 | char b[BDEVNAME_SIZE]; |
1241 | ||
1242 | if (likely(q)) | |
1243 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1244 | else | |
1245 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1246 | dm_device_name(t->md), | |
1247 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1248 | } |
1249 | ||
1250 | return r; | |
1251 | } | |
1252 | ||
cec47e3d KU |
1253 | int dm_table_any_busy_target(struct dm_table *t) |
1254 | { | |
1255 | unsigned i; | |
1256 | struct dm_target *ti; | |
1257 | ||
1258 | for (i = 0; i < t->num_targets; i++) { | |
1259 | ti = t->targets + i; | |
1260 | if (ti->type->busy && ti->type->busy(ti)) | |
1261 | return 1; | |
1262 | } | |
1263 | ||
1264 | return 0; | |
1265 | } | |
1266 | ||
1da177e4 LT |
1267 | void dm_table_unplug_all(struct dm_table *t) |
1268 | { | |
82b1519b | 1269 | struct dm_dev_internal *dd; |
afb24528 | 1270 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 | 1271 | |
afb24528 | 1272 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1273 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1274 | char b[BDEVNAME_SIZE]; |
1275 | ||
1276 | if (likely(q)) | |
1277 | blk_unplug(q); | |
1278 | else | |
1279 | DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s", | |
1280 | dm_device_name(t->md), | |
1281 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1282 | } |
1283 | } | |
1284 | ||
1134e5ae MA |
1285 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1286 | { | |
1134e5ae MA |
1287 | return t->md; |
1288 | } | |
1289 | ||
5ae89a87 MS |
1290 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1291 | sector_t start, sector_t len, void *data) | |
1292 | { | |
1293 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1294 | ||
1295 | return q && blk_queue_discard(q); | |
1296 | } | |
1297 | ||
1298 | bool dm_table_supports_discards(struct dm_table *t) | |
1299 | { | |
1300 | struct dm_target *ti; | |
1301 | unsigned i = 0; | |
1302 | ||
1303 | if (!t->discards_supported) | |
1304 | return 0; | |
1305 | ||
1306 | /* | |
1307 | * Ensure that at least one underlying device supports discards. | |
1308 | * t->devices includes internal dm devices such as mirror logs | |
1309 | * so we need to use iterate_devices here, which targets | |
1310 | * supporting discard must provide. | |
1311 | */ | |
1312 | while (i < dm_table_get_num_targets(t)) { | |
1313 | ti = dm_table_get_target(t, i++); | |
1314 | ||
1315 | if (ti->type->iterate_devices && | |
1316 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1317 | return 1; | |
1318 | } | |
1319 | ||
1320 | return 0; | |
1321 | } | |
1322 | ||
1da177e4 LT |
1323 | EXPORT_SYMBOL(dm_vcalloc); |
1324 | EXPORT_SYMBOL(dm_get_device); | |
1325 | EXPORT_SYMBOL(dm_put_device); | |
1326 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1327 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1328 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1329 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1330 | EXPORT_SYMBOL(dm_table_put); |
1331 | EXPORT_SYMBOL(dm_table_get); | |
1332 | EXPORT_SYMBOL(dm_table_unplug_all); |