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) | |
72d4cd9f MS |
520 | blk_limits_max_hw_sectors(limits, |
521 | (unsigned int) (PAGE_SIZE >> 9)); | |
754c5fc7 | 522 | return 0; |
3cb40214 BR |
523 | } |
524 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 525 | |
8215d6ec NK |
526 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
527 | struct dm_dev **result) | |
3cb40214 | 528 | { |
8215d6ec | 529 | return __table_get_device(ti->table, ti, path, mode, result); |
1da177e4 LT |
530 | } |
531 | ||
754c5fc7 | 532 | |
1da177e4 LT |
533 | /* |
534 | * Decrement a devices use count and remove it if necessary. | |
535 | */ | |
82b1519b | 536 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 537 | { |
82b1519b MP |
538 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
539 | dm_dev); | |
540 | ||
1da177e4 | 541 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 542 | close_dev(dd, ti->table->md); |
1da177e4 LT |
543 | list_del(&dd->list); |
544 | kfree(dd); | |
545 | } | |
546 | } | |
547 | ||
548 | /* | |
549 | * Checks to see if the target joins onto the end of the table. | |
550 | */ | |
551 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
552 | { | |
553 | struct dm_target *prev; | |
554 | ||
555 | if (!table->num_targets) | |
556 | return !ti->begin; | |
557 | ||
558 | prev = &table->targets[table->num_targets - 1]; | |
559 | return (ti->begin == (prev->begin + prev->len)); | |
560 | } | |
561 | ||
562 | /* | |
563 | * Used to dynamically allocate the arg array. | |
564 | */ | |
565 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
566 | { | |
567 | char **argv; | |
568 | unsigned new_size; | |
569 | ||
570 | new_size = *array_size ? *array_size * 2 : 64; | |
571 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
572 | if (argv) { | |
573 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
574 | *array_size = new_size; | |
575 | } | |
576 | ||
577 | kfree(old_argv); | |
578 | return argv; | |
579 | } | |
580 | ||
581 | /* | |
582 | * Destructively splits up the argument list to pass to ctr. | |
583 | */ | |
584 | int dm_split_args(int *argc, char ***argvp, char *input) | |
585 | { | |
586 | char *start, *end = input, *out, **argv = NULL; | |
587 | unsigned array_size = 0; | |
588 | ||
589 | *argc = 0; | |
814d6862 DT |
590 | |
591 | if (!input) { | |
592 | *argvp = NULL; | |
593 | return 0; | |
594 | } | |
595 | ||
1da177e4 LT |
596 | argv = realloc_argv(&array_size, argv); |
597 | if (!argv) | |
598 | return -ENOMEM; | |
599 | ||
600 | while (1) { | |
1da177e4 | 601 | /* Skip whitespace */ |
e7d2860b | 602 | start = skip_spaces(end); |
1da177e4 LT |
603 | |
604 | if (!*start) | |
605 | break; /* success, we hit the end */ | |
606 | ||
607 | /* 'out' is used to remove any back-quotes */ | |
608 | end = out = start; | |
609 | while (*end) { | |
610 | /* Everything apart from '\0' can be quoted */ | |
611 | if (*end == '\\' && *(end + 1)) { | |
612 | *out++ = *(end + 1); | |
613 | end += 2; | |
614 | continue; | |
615 | } | |
616 | ||
617 | if (isspace(*end)) | |
618 | break; /* end of token */ | |
619 | ||
620 | *out++ = *end++; | |
621 | } | |
622 | ||
623 | /* have we already filled the array ? */ | |
624 | if ((*argc + 1) > array_size) { | |
625 | argv = realloc_argv(&array_size, argv); | |
626 | if (!argv) | |
627 | return -ENOMEM; | |
628 | } | |
629 | ||
630 | /* we know this is whitespace */ | |
631 | if (*end) | |
632 | end++; | |
633 | ||
634 | /* terminate the string and put it in the array */ | |
635 | *out = '\0'; | |
636 | argv[*argc] = start; | |
637 | (*argc)++; | |
638 | } | |
639 | ||
640 | *argvp = argv; | |
641 | return 0; | |
642 | } | |
643 | ||
be6d4305 MS |
644 | /* |
645 | * Impose necessary and sufficient conditions on a devices's table such | |
646 | * that any incoming bio which respects its logical_block_size can be | |
647 | * processed successfully. If it falls across the boundary between | |
648 | * two or more targets, the size of each piece it gets split into must | |
649 | * be compatible with the logical_block_size of the target processing it. | |
650 | */ | |
754c5fc7 MS |
651 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
652 | struct queue_limits *limits) | |
be6d4305 MS |
653 | { |
654 | /* | |
655 | * This function uses arithmetic modulo the logical_block_size | |
656 | * (in units of 512-byte sectors). | |
657 | */ | |
658 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 659 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
660 | |
661 | /* | |
662 | * Offset of the start of the next table entry, mod logical_block_size. | |
663 | */ | |
664 | unsigned short next_target_start = 0; | |
665 | ||
666 | /* | |
667 | * Given an aligned bio that extends beyond the end of a | |
668 | * target, how many sectors must the next target handle? | |
669 | */ | |
670 | unsigned short remaining = 0; | |
671 | ||
672 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 673 | struct queue_limits ti_limits; |
be6d4305 MS |
674 | unsigned i = 0; |
675 | ||
676 | /* | |
677 | * Check each entry in the table in turn. | |
678 | */ | |
679 | while (i < dm_table_get_num_targets(table)) { | |
680 | ti = dm_table_get_target(table, i++); | |
681 | ||
754c5fc7 MS |
682 | blk_set_default_limits(&ti_limits); |
683 | ||
684 | /* combine all target devices' limits */ | |
685 | if (ti->type->iterate_devices) | |
686 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
687 | &ti_limits); | |
688 | ||
be6d4305 MS |
689 | /* |
690 | * If the remaining sectors fall entirely within this | |
691 | * table entry are they compatible with its logical_block_size? | |
692 | */ | |
693 | if (remaining < ti->len && | |
754c5fc7 | 694 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
695 | SECTOR_SHIFT) - 1)) |
696 | break; /* Error */ | |
697 | ||
698 | next_target_start = | |
699 | (unsigned short) ((next_target_start + ti->len) & | |
700 | (device_logical_block_size_sects - 1)); | |
701 | remaining = next_target_start ? | |
702 | device_logical_block_size_sects - next_target_start : 0; | |
703 | } | |
704 | ||
705 | if (remaining) { | |
706 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 707 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
708 | dm_device_name(table->md), i, |
709 | (unsigned long long) ti->begin, | |
710 | (unsigned long long) ti->len, | |
754c5fc7 | 711 | limits->logical_block_size); |
be6d4305 MS |
712 | return -EINVAL; |
713 | } | |
714 | ||
715 | return 0; | |
716 | } | |
717 | ||
1da177e4 LT |
718 | int dm_table_add_target(struct dm_table *t, const char *type, |
719 | sector_t start, sector_t len, char *params) | |
720 | { | |
721 | int r = -EINVAL, argc; | |
722 | char **argv; | |
723 | struct dm_target *tgt; | |
724 | ||
725 | if ((r = check_space(t))) | |
726 | return r; | |
727 | ||
728 | tgt = t->targets + t->num_targets; | |
729 | memset(tgt, 0, sizeof(*tgt)); | |
730 | ||
731 | if (!len) { | |
72d94861 | 732 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
733 | return -EINVAL; |
734 | } | |
735 | ||
736 | tgt->type = dm_get_target_type(type); | |
737 | if (!tgt->type) { | |
72d94861 AK |
738 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
739 | type); | |
1da177e4 LT |
740 | return -EINVAL; |
741 | } | |
742 | ||
743 | tgt->table = t; | |
744 | tgt->begin = start; | |
745 | tgt->len = len; | |
746 | tgt->error = "Unknown error"; | |
747 | ||
748 | /* | |
749 | * Does this target adjoin the previous one ? | |
750 | */ | |
751 | if (!adjoin(t, tgt)) { | |
752 | tgt->error = "Gap in table"; | |
753 | r = -EINVAL; | |
754 | goto bad; | |
755 | } | |
756 | ||
757 | r = dm_split_args(&argc, &argv, params); | |
758 | if (r) { | |
759 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
760 | goto bad; | |
761 | } | |
762 | ||
763 | r = tgt->type->ctr(tgt, argc, argv); | |
764 | kfree(argv); | |
765 | if (r) | |
766 | goto bad; | |
767 | ||
768 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
769 | ||
5ae89a87 MS |
770 | if (!tgt->num_discard_requests) |
771 | t->discards_supported = 0; | |
772 | ||
1da177e4 LT |
773 | return 0; |
774 | ||
775 | bad: | |
72d94861 | 776 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
777 | dm_put_target_type(tgt->type); |
778 | return r; | |
779 | } | |
780 | ||
26803b9f | 781 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
782 | { |
783 | unsigned i; | |
784 | unsigned bio_based = 0, request_based = 0; | |
785 | struct dm_target *tgt; | |
786 | struct dm_dev_internal *dd; | |
787 | struct list_head *devices; | |
788 | ||
789 | for (i = 0; i < t->num_targets; i++) { | |
790 | tgt = t->targets + i; | |
791 | if (dm_target_request_based(tgt)) | |
792 | request_based = 1; | |
793 | else | |
794 | bio_based = 1; | |
795 | ||
796 | if (bio_based && request_based) { | |
797 | DMWARN("Inconsistent table: different target types" | |
798 | " can't be mixed up"); | |
799 | return -EINVAL; | |
800 | } | |
801 | } | |
802 | ||
803 | if (bio_based) { | |
804 | /* We must use this table as bio-based */ | |
805 | t->type = DM_TYPE_BIO_BASED; | |
806 | return 0; | |
807 | } | |
808 | ||
809 | BUG_ON(!request_based); /* No targets in this table */ | |
810 | ||
811 | /* Non-request-stackable devices can't be used for request-based dm */ | |
812 | devices = dm_table_get_devices(t); | |
813 | list_for_each_entry(dd, devices, list) { | |
814 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
815 | DMWARN("table load rejected: including" | |
816 | " non-request-stackable devices"); | |
817 | return -EINVAL; | |
818 | } | |
819 | } | |
820 | ||
821 | /* | |
822 | * Request-based dm supports only tables that have a single target now. | |
823 | * To support multiple targets, request splitting support is needed, | |
824 | * and that needs lots of changes in the block-layer. | |
825 | * (e.g. request completion process for partial completion.) | |
826 | */ | |
827 | if (t->num_targets > 1) { | |
828 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
829 | return -EINVAL; | |
830 | } | |
831 | ||
832 | t->type = DM_TYPE_REQUEST_BASED; | |
833 | ||
834 | return 0; | |
835 | } | |
836 | ||
837 | unsigned dm_table_get_type(struct dm_table *t) | |
838 | { | |
839 | return t->type; | |
840 | } | |
841 | ||
842 | bool dm_table_request_based(struct dm_table *t) | |
843 | { | |
844 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
845 | } | |
846 | ||
847 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
848 | { | |
849 | unsigned type = dm_table_get_type(t); | |
850 | ||
851 | if (unlikely(type == DM_TYPE_NONE)) { | |
852 | DMWARN("no table type is set, can't allocate mempools"); | |
853 | return -EINVAL; | |
854 | } | |
855 | ||
856 | t->mempools = dm_alloc_md_mempools(type); | |
857 | if (!t->mempools) | |
858 | return -ENOMEM; | |
859 | ||
860 | return 0; | |
861 | } | |
862 | ||
863 | void dm_table_free_md_mempools(struct dm_table *t) | |
864 | { | |
865 | dm_free_md_mempools(t->mempools); | |
866 | t->mempools = NULL; | |
867 | } | |
868 | ||
869 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
870 | { | |
871 | return t->mempools; | |
872 | } | |
873 | ||
1da177e4 LT |
874 | static int setup_indexes(struct dm_table *t) |
875 | { | |
876 | int i; | |
877 | unsigned int total = 0; | |
878 | sector_t *indexes; | |
879 | ||
880 | /* allocate the space for *all* the indexes */ | |
881 | for (i = t->depth - 2; i >= 0; i--) { | |
882 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
883 | total += t->counts[i]; | |
884 | } | |
885 | ||
886 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
887 | if (!indexes) | |
888 | return -ENOMEM; | |
889 | ||
890 | /* set up internal nodes, bottom-up */ | |
82d601dc | 891 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
892 | t->index[i] = indexes; |
893 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
894 | setup_btree_index(i, t); | |
895 | } | |
896 | ||
897 | return 0; | |
898 | } | |
899 | ||
900 | /* | |
901 | * Builds the btree to index the map. | |
902 | */ | |
26803b9f | 903 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
904 | { |
905 | int r = 0; | |
906 | unsigned int leaf_nodes; | |
907 | ||
1da177e4 LT |
908 | /* how many indexes will the btree have ? */ |
909 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
910 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
911 | ||
912 | /* leaf layer has already been set up */ | |
913 | t->counts[t->depth - 1] = leaf_nodes; | |
914 | t->index[t->depth - 1] = t->highs; | |
915 | ||
916 | if (t->depth >= 2) | |
917 | r = setup_indexes(t); | |
918 | ||
919 | return r; | |
920 | } | |
921 | ||
26803b9f WD |
922 | /* |
923 | * Register the mapped device for blk_integrity support if | |
924 | * the underlying devices support it. | |
925 | */ | |
926 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
927 | { | |
928 | struct list_head *devices = dm_table_get_devices(t); | |
929 | struct dm_dev_internal *dd; | |
930 | ||
931 | list_for_each_entry(dd, devices, list) | |
932 | if (bdev_get_integrity(dd->dm_dev.bdev)) | |
933 | return blk_integrity_register(dm_disk(md), NULL); | |
934 | ||
935 | return 0; | |
936 | } | |
937 | ||
938 | /* | |
939 | * Prepares the table for use by building the indices, | |
940 | * setting the type, and allocating mempools. | |
941 | */ | |
942 | int dm_table_complete(struct dm_table *t) | |
943 | { | |
944 | int r; | |
945 | ||
946 | r = dm_table_set_type(t); | |
947 | if (r) { | |
948 | DMERR("unable to set table type"); | |
949 | return r; | |
950 | } | |
951 | ||
952 | r = dm_table_build_index(t); | |
953 | if (r) { | |
954 | DMERR("unable to build btrees"); | |
955 | return r; | |
956 | } | |
957 | ||
958 | r = dm_table_prealloc_integrity(t, t->md); | |
959 | if (r) { | |
960 | DMERR("could not register integrity profile."); | |
961 | return r; | |
962 | } | |
963 | ||
964 | r = dm_table_alloc_md_mempools(t); | |
965 | if (r) | |
966 | DMERR("unable to allocate mempools"); | |
967 | ||
968 | return r; | |
969 | } | |
970 | ||
48c9c27b | 971 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
972 | void dm_table_event_callback(struct dm_table *t, |
973 | void (*fn)(void *), void *context) | |
974 | { | |
48c9c27b | 975 | mutex_lock(&_event_lock); |
1da177e4 LT |
976 | t->event_fn = fn; |
977 | t->event_context = context; | |
48c9c27b | 978 | mutex_unlock(&_event_lock); |
1da177e4 LT |
979 | } |
980 | ||
981 | void dm_table_event(struct dm_table *t) | |
982 | { | |
983 | /* | |
984 | * You can no longer call dm_table_event() from interrupt | |
985 | * context, use a bottom half instead. | |
986 | */ | |
987 | BUG_ON(in_interrupt()); | |
988 | ||
48c9c27b | 989 | mutex_lock(&_event_lock); |
1da177e4 LT |
990 | if (t->event_fn) |
991 | t->event_fn(t->event_context); | |
48c9c27b | 992 | mutex_unlock(&_event_lock); |
1da177e4 LT |
993 | } |
994 | ||
995 | sector_t dm_table_get_size(struct dm_table *t) | |
996 | { | |
997 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
998 | } | |
999 | ||
1000 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1001 | { | |
14353539 | 1002 | if (index >= t->num_targets) |
1da177e4 LT |
1003 | return NULL; |
1004 | ||
1005 | return t->targets + index; | |
1006 | } | |
1007 | ||
1008 | /* | |
1009 | * Search the btree for the correct target. | |
512875bd JN |
1010 | * |
1011 | * Caller should check returned pointer with dm_target_is_valid() | |
1012 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1013 | */ |
1014 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1015 | { | |
1016 | unsigned int l, n = 0, k = 0; | |
1017 | sector_t *node; | |
1018 | ||
1019 | for (l = 0; l < t->depth; l++) { | |
1020 | n = get_child(n, k); | |
1021 | node = get_node(t, l, n); | |
1022 | ||
1023 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1024 | if (node[k] >= sector) | |
1025 | break; | |
1026 | } | |
1027 | ||
1028 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1029 | } | |
1030 | ||
754c5fc7 MS |
1031 | /* |
1032 | * Establish the new table's queue_limits and validate them. | |
1033 | */ | |
1034 | int dm_calculate_queue_limits(struct dm_table *table, | |
1035 | struct queue_limits *limits) | |
1036 | { | |
1037 | struct dm_target *uninitialized_var(ti); | |
1038 | struct queue_limits ti_limits; | |
1039 | unsigned i = 0; | |
1040 | ||
1041 | blk_set_default_limits(limits); | |
1042 | ||
1043 | while (i < dm_table_get_num_targets(table)) { | |
1044 | blk_set_default_limits(&ti_limits); | |
1045 | ||
1046 | ti = dm_table_get_target(table, i++); | |
1047 | ||
1048 | if (!ti->type->iterate_devices) | |
1049 | goto combine_limits; | |
1050 | ||
1051 | /* | |
1052 | * Combine queue limits of all the devices this target uses. | |
1053 | */ | |
1054 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1055 | &ti_limits); | |
1056 | ||
40bea431 MS |
1057 | /* Set I/O hints portion of queue limits */ |
1058 | if (ti->type->io_hints) | |
1059 | ti->type->io_hints(ti, &ti_limits); | |
1060 | ||
754c5fc7 MS |
1061 | /* |
1062 | * Check each device area is consistent with the target's | |
1063 | * overall queue limits. | |
1064 | */ | |
f6a1ed10 MP |
1065 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1066 | &ti_limits)) | |
754c5fc7 MS |
1067 | return -EINVAL; |
1068 | ||
1069 | combine_limits: | |
1070 | /* | |
1071 | * Merge this target's queue limits into the overall limits | |
1072 | * for the table. | |
1073 | */ | |
1074 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1075 | DMWARN("%s: adding target device " |
754c5fc7 | 1076 | "(start sect %llu len %llu) " |
b27d7f16 | 1077 | "caused an alignment inconsistency", |
754c5fc7 MS |
1078 | dm_device_name(table->md), |
1079 | (unsigned long long) ti->begin, | |
1080 | (unsigned long long) ti->len); | |
1081 | } | |
1082 | ||
1083 | return validate_hardware_logical_block_alignment(table, limits); | |
1084 | } | |
1085 | ||
9c47008d MP |
1086 | /* |
1087 | * Set the integrity profile for this device if all devices used have | |
1088 | * matching profiles. | |
1089 | */ | |
1090 | static void dm_table_set_integrity(struct dm_table *t) | |
1091 | { | |
1092 | struct list_head *devices = dm_table_get_devices(t); | |
1093 | struct dm_dev_internal *prev = NULL, *dd = NULL; | |
1094 | ||
1095 | if (!blk_get_integrity(dm_disk(t->md))) | |
1096 | return; | |
1097 | ||
1098 | list_for_each_entry(dd, devices, list) { | |
1099 | if (prev && | |
1100 | blk_integrity_compare(prev->dm_dev.bdev->bd_disk, | |
1101 | dd->dm_dev.bdev->bd_disk) < 0) { | |
1102 | DMWARN("%s: integrity not set: %s and %s mismatch", | |
1103 | dm_device_name(t->md), | |
1104 | prev->dm_dev.bdev->bd_disk->disk_name, | |
1105 | dd->dm_dev.bdev->bd_disk->disk_name); | |
1106 | goto no_integrity; | |
1107 | } | |
1108 | prev = dd; | |
1109 | } | |
1110 | ||
1111 | if (!prev || !bdev_get_integrity(prev->dm_dev.bdev)) | |
1112 | goto no_integrity; | |
1113 | ||
1114 | blk_integrity_register(dm_disk(t->md), | |
1115 | bdev_get_integrity(prev->dm_dev.bdev)); | |
1116 | ||
1117 | return; | |
1118 | ||
1119 | no_integrity: | |
1120 | blk_integrity_register(dm_disk(t->md), NULL); | |
1121 | ||
1122 | return; | |
1123 | } | |
1124 | ||
754c5fc7 MS |
1125 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1126 | struct queue_limits *limits) | |
1da177e4 LT |
1127 | { |
1128 | /* | |
1197764e | 1129 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1130 | */ |
754c5fc7 | 1131 | q->limits = *limits; |
c9a3f6d6 | 1132 | |
5ae89a87 MS |
1133 | if (!dm_table_supports_discards(t)) |
1134 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1135 | else | |
1136 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1137 | ||
9c47008d | 1138 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1139 | |
1140 | /* | |
1141 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1142 | * visible to other CPUs because, once the flag is set, incoming bios | |
1143 | * are processed by request-based dm, which refers to the queue | |
1144 | * settings. | |
1145 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1146 | * md->deferred where queue settings are not needed yet. | |
1147 | * Those bios are passed to request-based dm at the resume time. | |
1148 | */ | |
1149 | smp_mb(); | |
1150 | if (dm_table_request_based(t)) | |
1151 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1152 | } |
1153 | ||
1154 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1155 | { | |
1156 | return t->num_targets; | |
1157 | } | |
1158 | ||
1159 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1160 | { | |
1161 | return &t->devices; | |
1162 | } | |
1163 | ||
aeb5d727 | 1164 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1165 | { |
1166 | return t->mode; | |
1167 | } | |
1168 | ||
1169 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1170 | { | |
1171 | int i = t->num_targets; | |
1172 | struct dm_target *ti = t->targets; | |
1173 | ||
1174 | while (i--) { | |
1175 | if (postsuspend) { | |
1176 | if (ti->type->postsuspend) | |
1177 | ti->type->postsuspend(ti); | |
1178 | } else if (ti->type->presuspend) | |
1179 | ti->type->presuspend(ti); | |
1180 | ||
1181 | ti++; | |
1182 | } | |
1183 | } | |
1184 | ||
1185 | void dm_table_presuspend_targets(struct dm_table *t) | |
1186 | { | |
cf222b37 AK |
1187 | if (!t) |
1188 | return; | |
1189 | ||
e8488d08 | 1190 | suspend_targets(t, 0); |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1194 | { | |
cf222b37 AK |
1195 | if (!t) |
1196 | return; | |
1197 | ||
e8488d08 | 1198 | suspend_targets(t, 1); |
1da177e4 LT |
1199 | } |
1200 | ||
8757b776 | 1201 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1202 | { |
8757b776 MB |
1203 | int i, r = 0; |
1204 | ||
1205 | for (i = 0; i < t->num_targets; i++) { | |
1206 | struct dm_target *ti = t->targets + i; | |
1207 | ||
1208 | if (!ti->type->preresume) | |
1209 | continue; | |
1210 | ||
1211 | r = ti->type->preresume(ti); | |
1212 | if (r) | |
1213 | return r; | |
1214 | } | |
1da177e4 LT |
1215 | |
1216 | for (i = 0; i < t->num_targets; i++) { | |
1217 | struct dm_target *ti = t->targets + i; | |
1218 | ||
1219 | if (ti->type->resume) | |
1220 | ti->type->resume(ti); | |
1221 | } | |
8757b776 MB |
1222 | |
1223 | return 0; | |
1da177e4 LT |
1224 | } |
1225 | ||
1226 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | |
1227 | { | |
82b1519b | 1228 | struct dm_dev_internal *dd; |
afb24528 | 1229 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 LT |
1230 | int r = 0; |
1231 | ||
afb24528 | 1232 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1233 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1234 | char b[BDEVNAME_SIZE]; |
1235 | ||
1236 | if (likely(q)) | |
1237 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1238 | else | |
1239 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1240 | dm_device_name(t->md), | |
1241 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1242 | } |
1243 | ||
1244 | return r; | |
1245 | } | |
1246 | ||
cec47e3d KU |
1247 | int dm_table_any_busy_target(struct dm_table *t) |
1248 | { | |
1249 | unsigned i; | |
1250 | struct dm_target *ti; | |
1251 | ||
1252 | for (i = 0; i < t->num_targets; i++) { | |
1253 | ti = t->targets + i; | |
1254 | if (ti->type->busy && ti->type->busy(ti)) | |
1255 | return 1; | |
1256 | } | |
1257 | ||
1258 | return 0; | |
1259 | } | |
1260 | ||
1da177e4 LT |
1261 | void dm_table_unplug_all(struct dm_table *t) |
1262 | { | |
82b1519b | 1263 | struct dm_dev_internal *dd; |
afb24528 | 1264 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 | 1265 | |
afb24528 | 1266 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1267 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1268 | char b[BDEVNAME_SIZE]; |
1269 | ||
1270 | if (likely(q)) | |
1271 | blk_unplug(q); | |
1272 | else | |
1273 | DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s", | |
1274 | dm_device_name(t->md), | |
1275 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1276 | } |
1277 | } | |
1278 | ||
1134e5ae MA |
1279 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1280 | { | |
1134e5ae MA |
1281 | return t->md; |
1282 | } | |
1283 | ||
5ae89a87 MS |
1284 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1285 | sector_t start, sector_t len, void *data) | |
1286 | { | |
1287 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1288 | ||
1289 | return q && blk_queue_discard(q); | |
1290 | } | |
1291 | ||
1292 | bool dm_table_supports_discards(struct dm_table *t) | |
1293 | { | |
1294 | struct dm_target *ti; | |
1295 | unsigned i = 0; | |
1296 | ||
1297 | if (!t->discards_supported) | |
1298 | return 0; | |
1299 | ||
1300 | /* | |
1301 | * Ensure that at least one underlying device supports discards. | |
1302 | * t->devices includes internal dm devices such as mirror logs | |
1303 | * so we need to use iterate_devices here, which targets | |
1304 | * supporting discard must provide. | |
1305 | */ | |
1306 | while (i < dm_table_get_num_targets(t)) { | |
1307 | ti = dm_table_get_target(t, i++); | |
1308 | ||
1309 | if (ti->type->iterate_devices && | |
1310 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1311 | return 1; | |
1312 | } | |
1313 | ||
1314 | return 0; | |
1315 | } | |
1316 | ||
1da177e4 LT |
1317 | EXPORT_SYMBOL(dm_vcalloc); |
1318 | EXPORT_SYMBOL(dm_get_device); | |
1319 | EXPORT_SYMBOL(dm_put_device); | |
1320 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1321 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1322 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1323 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1324 | EXPORT_SYMBOL(dm_table_put); |
1325 | EXPORT_SYMBOL(dm_table_get); | |
1326 | EXPORT_SYMBOL(dm_table_unplug_all); |