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 | |
d4d77629 | 328 | bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr); |
1da177e4 LT |
329 | if (IS_ERR(bdev)) |
330 | return PTR_ERR(bdev); | |
e09b457b | 331 | |
e09b457b TH |
332 | r = bd_link_disk_holder(bdev, dm_disk(md)); |
333 | if (r) { | |
e525fd89 | 334 | blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL); |
e09b457b TH |
335 | return r; |
336 | } | |
337 | ||
338 | d->dm_dev.bdev = bdev; | |
339 | return 0; | |
1da177e4 LT |
340 | } |
341 | ||
342 | /* | |
343 | * Close a device that we've been using. | |
344 | */ | |
82b1519b | 345 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 346 | { |
82b1519b | 347 | if (!d->dm_dev.bdev) |
1da177e4 LT |
348 | return; |
349 | ||
e525fd89 | 350 | blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL); |
82b1519b | 351 | d->dm_dev.bdev = NULL; |
1da177e4 LT |
352 | } |
353 | ||
354 | /* | |
f6a1ed10 | 355 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 356 | */ |
f6a1ed10 MP |
357 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
358 | sector_t start, sector_t len, void *data) | |
1da177e4 | 359 | { |
754c5fc7 MS |
360 | struct queue_limits *limits = data; |
361 | struct block_device *bdev = dev->bdev; | |
362 | sector_t dev_size = | |
363 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 364 | unsigned short logical_block_size_sectors = |
754c5fc7 | 365 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 366 | char b[BDEVNAME_SIZE]; |
2cd54d9b MA |
367 | |
368 | if (!dev_size) | |
f6a1ed10 | 369 | return 0; |
2cd54d9b | 370 | |
5dea271b | 371 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
372 | DMWARN("%s: %s too small for target: " |
373 | "start=%llu, len=%llu, dev_size=%llu", | |
374 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
375 | (unsigned long long)start, | |
376 | (unsigned long long)len, | |
377 | (unsigned long long)dev_size); | |
f6a1ed10 | 378 | return 1; |
02acc3a4 MS |
379 | } |
380 | ||
381 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 382 | return 0; |
02acc3a4 MS |
383 | |
384 | if (start & (logical_block_size_sectors - 1)) { | |
385 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 386 | "logical block size %u of %s", |
02acc3a4 MS |
387 | dm_device_name(ti->table->md), |
388 | (unsigned long long)start, | |
754c5fc7 | 389 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 390 | return 1; |
02acc3a4 MS |
391 | } |
392 | ||
5dea271b | 393 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 394 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 395 | "logical block size %u of %s", |
02acc3a4 | 396 | dm_device_name(ti->table->md), |
5dea271b | 397 | (unsigned long long)len, |
754c5fc7 | 398 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 399 | return 1; |
02acc3a4 MS |
400 | } |
401 | ||
f6a1ed10 | 402 | return 0; |
1da177e4 LT |
403 | } |
404 | ||
405 | /* | |
570b9d96 | 406 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 407 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
408 | * device and not to touch the existing bdev field in case |
409 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 410 | */ |
aeb5d727 | 411 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 412 | struct mapped_device *md) |
1da177e4 LT |
413 | { |
414 | int r; | |
570b9d96 | 415 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 416 | |
570b9d96 AK |
417 | dd_new = dd_old = *dd; |
418 | ||
419 | dd_new.dm_dev.mode |= new_mode; | |
420 | dd_new.dm_dev.bdev = NULL; | |
421 | ||
422 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
423 | if (r) | |
424 | return r; | |
1da177e4 | 425 | |
82b1519b | 426 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 427 | close_dev(&dd_old, md); |
1da177e4 | 428 | |
570b9d96 | 429 | return 0; |
1da177e4 LT |
430 | } |
431 | ||
432 | /* | |
433 | * Add a device to the list, or just increment the usage count if | |
434 | * it's already present. | |
435 | */ | |
436 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
8215d6ec | 437 | const char *path, fmode_t mode, struct dm_dev **result) |
1da177e4 LT |
438 | { |
439 | int r; | |
69a2ce72 | 440 | dev_t uninitialized_var(dev); |
82b1519b | 441 | struct dm_dev_internal *dd; |
1da177e4 LT |
442 | unsigned int major, minor; |
443 | ||
547bc926 | 444 | BUG_ON(!t); |
1da177e4 LT |
445 | |
446 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
447 | /* Extract the major/minor numbers */ | |
448 | dev = MKDEV(major, minor); | |
449 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
450 | return -EOVERFLOW; | |
451 | } else { | |
452 | /* convert the path to a device */ | |
72e8264e CH |
453 | struct block_device *bdev = lookup_bdev(path); |
454 | ||
455 | if (IS_ERR(bdev)) | |
456 | return PTR_ERR(bdev); | |
457 | dev = bdev->bd_dev; | |
458 | bdput(bdev); | |
1da177e4 LT |
459 | } |
460 | ||
461 | dd = find_device(&t->devices, dev); | |
462 | if (!dd) { | |
463 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
464 | if (!dd) | |
465 | return -ENOMEM; | |
466 | ||
82b1519b MP |
467 | dd->dm_dev.mode = mode; |
468 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 469 | |
f165921d | 470 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
471 | kfree(dd); |
472 | return r; | |
473 | } | |
474 | ||
82b1519b | 475 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
476 | |
477 | atomic_set(&dd->count, 0); | |
478 | list_add(&dd->list, &t->devices); | |
479 | ||
82b1519b | 480 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 481 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
482 | if (r) |
483 | return r; | |
484 | } | |
485 | atomic_inc(&dd->count); | |
486 | ||
82b1519b | 487 | *result = &dd->dm_dev; |
1da177e4 LT |
488 | return 0; |
489 | } | |
490 | ||
754c5fc7 | 491 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 492 | sector_t start, sector_t len, void *data) |
1da177e4 | 493 | { |
754c5fc7 MS |
494 | struct queue_limits *limits = data; |
495 | struct block_device *bdev = dev->bdev; | |
165125e1 | 496 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
497 | char b[BDEVNAME_SIZE]; |
498 | ||
499 | if (unlikely(!q)) { | |
500 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
501 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 502 | return 0; |
0c2322e4 | 503 | } |
3cb40214 | 504 | |
b27d7f16 MP |
505 | if (bdev_stack_limits(limits, bdev, start) < 0) |
506 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
507 | "physical_block_size=%u, logical_block_size=%u, " |
508 | "alignment_offset=%u, start=%llu", | |
509 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
510 | q->limits.physical_block_size, | |
511 | q->limits.logical_block_size, | |
512 | q->limits.alignment_offset, | |
b27d7f16 | 513 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 514 | |
9980c638 MB |
515 | /* |
516 | * Check if merge fn is supported. | |
517 | * If not we'll force DM to use PAGE_SIZE or | |
518 | * smaller I/O, just to be safe. | |
3cb40214 | 519 | */ |
9980c638 MB |
520 | |
521 | if (q->merge_bvec_fn && !ti->type->merge) | |
754c5fc7 MS |
522 | limits->max_sectors = |
523 | min_not_zero(limits->max_sectors, | |
3cb40214 | 524 | (unsigned int) (PAGE_SIZE >> 9)); |
754c5fc7 | 525 | return 0; |
3cb40214 BR |
526 | } |
527 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 528 | |
8215d6ec NK |
529 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
530 | struct dm_dev **result) | |
3cb40214 | 531 | { |
8215d6ec | 532 | return __table_get_device(ti->table, ti, path, mode, result); |
1da177e4 LT |
533 | } |
534 | ||
754c5fc7 | 535 | |
1da177e4 LT |
536 | /* |
537 | * Decrement a devices use count and remove it if necessary. | |
538 | */ | |
82b1519b | 539 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 540 | { |
82b1519b MP |
541 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
542 | dm_dev); | |
543 | ||
1da177e4 | 544 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 545 | close_dev(dd, ti->table->md); |
1da177e4 LT |
546 | list_del(&dd->list); |
547 | kfree(dd); | |
548 | } | |
549 | } | |
550 | ||
551 | /* | |
552 | * Checks to see if the target joins onto the end of the table. | |
553 | */ | |
554 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
555 | { | |
556 | struct dm_target *prev; | |
557 | ||
558 | if (!table->num_targets) | |
559 | return !ti->begin; | |
560 | ||
561 | prev = &table->targets[table->num_targets - 1]; | |
562 | return (ti->begin == (prev->begin + prev->len)); | |
563 | } | |
564 | ||
565 | /* | |
566 | * Used to dynamically allocate the arg array. | |
567 | */ | |
568 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
569 | { | |
570 | char **argv; | |
571 | unsigned new_size; | |
572 | ||
573 | new_size = *array_size ? *array_size * 2 : 64; | |
574 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
575 | if (argv) { | |
576 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
577 | *array_size = new_size; | |
578 | } | |
579 | ||
580 | kfree(old_argv); | |
581 | return argv; | |
582 | } | |
583 | ||
584 | /* | |
585 | * Destructively splits up the argument list to pass to ctr. | |
586 | */ | |
587 | int dm_split_args(int *argc, char ***argvp, char *input) | |
588 | { | |
589 | char *start, *end = input, *out, **argv = NULL; | |
590 | unsigned array_size = 0; | |
591 | ||
592 | *argc = 0; | |
814d6862 DT |
593 | |
594 | if (!input) { | |
595 | *argvp = NULL; | |
596 | return 0; | |
597 | } | |
598 | ||
1da177e4 LT |
599 | argv = realloc_argv(&array_size, argv); |
600 | if (!argv) | |
601 | return -ENOMEM; | |
602 | ||
603 | while (1) { | |
1da177e4 | 604 | /* Skip whitespace */ |
e7d2860b | 605 | start = skip_spaces(end); |
1da177e4 LT |
606 | |
607 | if (!*start) | |
608 | break; /* success, we hit the end */ | |
609 | ||
610 | /* 'out' is used to remove any back-quotes */ | |
611 | end = out = start; | |
612 | while (*end) { | |
613 | /* Everything apart from '\0' can be quoted */ | |
614 | if (*end == '\\' && *(end + 1)) { | |
615 | *out++ = *(end + 1); | |
616 | end += 2; | |
617 | continue; | |
618 | } | |
619 | ||
620 | if (isspace(*end)) | |
621 | break; /* end of token */ | |
622 | ||
623 | *out++ = *end++; | |
624 | } | |
625 | ||
626 | /* have we already filled the array ? */ | |
627 | if ((*argc + 1) > array_size) { | |
628 | argv = realloc_argv(&array_size, argv); | |
629 | if (!argv) | |
630 | return -ENOMEM; | |
631 | } | |
632 | ||
633 | /* we know this is whitespace */ | |
634 | if (*end) | |
635 | end++; | |
636 | ||
637 | /* terminate the string and put it in the array */ | |
638 | *out = '\0'; | |
639 | argv[*argc] = start; | |
640 | (*argc)++; | |
641 | } | |
642 | ||
643 | *argvp = argv; | |
644 | return 0; | |
645 | } | |
646 | ||
be6d4305 MS |
647 | /* |
648 | * Impose necessary and sufficient conditions on a devices's table such | |
649 | * that any incoming bio which respects its logical_block_size can be | |
650 | * processed successfully. If it falls across the boundary between | |
651 | * two or more targets, the size of each piece it gets split into must | |
652 | * be compatible with the logical_block_size of the target processing it. | |
653 | */ | |
754c5fc7 MS |
654 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
655 | struct queue_limits *limits) | |
be6d4305 MS |
656 | { |
657 | /* | |
658 | * This function uses arithmetic modulo the logical_block_size | |
659 | * (in units of 512-byte sectors). | |
660 | */ | |
661 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 662 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
663 | |
664 | /* | |
665 | * Offset of the start of the next table entry, mod logical_block_size. | |
666 | */ | |
667 | unsigned short next_target_start = 0; | |
668 | ||
669 | /* | |
670 | * Given an aligned bio that extends beyond the end of a | |
671 | * target, how many sectors must the next target handle? | |
672 | */ | |
673 | unsigned short remaining = 0; | |
674 | ||
675 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 676 | struct queue_limits ti_limits; |
be6d4305 MS |
677 | unsigned i = 0; |
678 | ||
679 | /* | |
680 | * Check each entry in the table in turn. | |
681 | */ | |
682 | while (i < dm_table_get_num_targets(table)) { | |
683 | ti = dm_table_get_target(table, i++); | |
684 | ||
754c5fc7 MS |
685 | blk_set_default_limits(&ti_limits); |
686 | ||
687 | /* combine all target devices' limits */ | |
688 | if (ti->type->iterate_devices) | |
689 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
690 | &ti_limits); | |
691 | ||
be6d4305 MS |
692 | /* |
693 | * If the remaining sectors fall entirely within this | |
694 | * table entry are they compatible with its logical_block_size? | |
695 | */ | |
696 | if (remaining < ti->len && | |
754c5fc7 | 697 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
698 | SECTOR_SHIFT) - 1)) |
699 | break; /* Error */ | |
700 | ||
701 | next_target_start = | |
702 | (unsigned short) ((next_target_start + ti->len) & | |
703 | (device_logical_block_size_sects - 1)); | |
704 | remaining = next_target_start ? | |
705 | device_logical_block_size_sects - next_target_start : 0; | |
706 | } | |
707 | ||
708 | if (remaining) { | |
709 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 710 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
711 | dm_device_name(table->md), i, |
712 | (unsigned long long) ti->begin, | |
713 | (unsigned long long) ti->len, | |
754c5fc7 | 714 | limits->logical_block_size); |
be6d4305 MS |
715 | return -EINVAL; |
716 | } | |
717 | ||
718 | return 0; | |
719 | } | |
720 | ||
1da177e4 LT |
721 | int dm_table_add_target(struct dm_table *t, const char *type, |
722 | sector_t start, sector_t len, char *params) | |
723 | { | |
724 | int r = -EINVAL, argc; | |
725 | char **argv; | |
726 | struct dm_target *tgt; | |
727 | ||
728 | if ((r = check_space(t))) | |
729 | return r; | |
730 | ||
731 | tgt = t->targets + t->num_targets; | |
732 | memset(tgt, 0, sizeof(*tgt)); | |
733 | ||
734 | if (!len) { | |
72d94861 | 735 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
736 | return -EINVAL; |
737 | } | |
738 | ||
739 | tgt->type = dm_get_target_type(type); | |
740 | if (!tgt->type) { | |
72d94861 AK |
741 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
742 | type); | |
1da177e4 LT |
743 | return -EINVAL; |
744 | } | |
745 | ||
746 | tgt->table = t; | |
747 | tgt->begin = start; | |
748 | tgt->len = len; | |
749 | tgt->error = "Unknown error"; | |
750 | ||
751 | /* | |
752 | * Does this target adjoin the previous one ? | |
753 | */ | |
754 | if (!adjoin(t, tgt)) { | |
755 | tgt->error = "Gap in table"; | |
756 | r = -EINVAL; | |
757 | goto bad; | |
758 | } | |
759 | ||
760 | r = dm_split_args(&argc, &argv, params); | |
761 | if (r) { | |
762 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
763 | goto bad; | |
764 | } | |
765 | ||
766 | r = tgt->type->ctr(tgt, argc, argv); | |
767 | kfree(argv); | |
768 | if (r) | |
769 | goto bad; | |
770 | ||
771 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
772 | ||
5ae89a87 MS |
773 | if (!tgt->num_discard_requests) |
774 | t->discards_supported = 0; | |
775 | ||
1da177e4 LT |
776 | return 0; |
777 | ||
778 | bad: | |
72d94861 | 779 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
780 | dm_put_target_type(tgt->type); |
781 | return r; | |
782 | } | |
783 | ||
26803b9f | 784 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
785 | { |
786 | unsigned i; | |
787 | unsigned bio_based = 0, request_based = 0; | |
788 | struct dm_target *tgt; | |
789 | struct dm_dev_internal *dd; | |
790 | struct list_head *devices; | |
791 | ||
792 | for (i = 0; i < t->num_targets; i++) { | |
793 | tgt = t->targets + i; | |
794 | if (dm_target_request_based(tgt)) | |
795 | request_based = 1; | |
796 | else | |
797 | bio_based = 1; | |
798 | ||
799 | if (bio_based && request_based) { | |
800 | DMWARN("Inconsistent table: different target types" | |
801 | " can't be mixed up"); | |
802 | return -EINVAL; | |
803 | } | |
804 | } | |
805 | ||
806 | if (bio_based) { | |
807 | /* We must use this table as bio-based */ | |
808 | t->type = DM_TYPE_BIO_BASED; | |
809 | return 0; | |
810 | } | |
811 | ||
812 | BUG_ON(!request_based); /* No targets in this table */ | |
813 | ||
814 | /* Non-request-stackable devices can't be used for request-based dm */ | |
815 | devices = dm_table_get_devices(t); | |
816 | list_for_each_entry(dd, devices, list) { | |
817 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
818 | DMWARN("table load rejected: including" | |
819 | " non-request-stackable devices"); | |
820 | return -EINVAL; | |
821 | } | |
822 | } | |
823 | ||
824 | /* | |
825 | * Request-based dm supports only tables that have a single target now. | |
826 | * To support multiple targets, request splitting support is needed, | |
827 | * and that needs lots of changes in the block-layer. | |
828 | * (e.g. request completion process for partial completion.) | |
829 | */ | |
830 | if (t->num_targets > 1) { | |
831 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
832 | return -EINVAL; | |
833 | } | |
834 | ||
835 | t->type = DM_TYPE_REQUEST_BASED; | |
836 | ||
837 | return 0; | |
838 | } | |
839 | ||
840 | unsigned dm_table_get_type(struct dm_table *t) | |
841 | { | |
842 | return t->type; | |
843 | } | |
844 | ||
845 | bool dm_table_request_based(struct dm_table *t) | |
846 | { | |
847 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
848 | } | |
849 | ||
850 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
851 | { | |
852 | unsigned type = dm_table_get_type(t); | |
853 | ||
854 | if (unlikely(type == DM_TYPE_NONE)) { | |
855 | DMWARN("no table type is set, can't allocate mempools"); | |
856 | return -EINVAL; | |
857 | } | |
858 | ||
859 | t->mempools = dm_alloc_md_mempools(type); | |
860 | if (!t->mempools) | |
861 | return -ENOMEM; | |
862 | ||
863 | return 0; | |
864 | } | |
865 | ||
866 | void dm_table_free_md_mempools(struct dm_table *t) | |
867 | { | |
868 | dm_free_md_mempools(t->mempools); | |
869 | t->mempools = NULL; | |
870 | } | |
871 | ||
872 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
873 | { | |
874 | return t->mempools; | |
875 | } | |
876 | ||
1da177e4 LT |
877 | static int setup_indexes(struct dm_table *t) |
878 | { | |
879 | int i; | |
880 | unsigned int total = 0; | |
881 | sector_t *indexes; | |
882 | ||
883 | /* allocate the space for *all* the indexes */ | |
884 | for (i = t->depth - 2; i >= 0; i--) { | |
885 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
886 | total += t->counts[i]; | |
887 | } | |
888 | ||
889 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
890 | if (!indexes) | |
891 | return -ENOMEM; | |
892 | ||
893 | /* set up internal nodes, bottom-up */ | |
82d601dc | 894 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
895 | t->index[i] = indexes; |
896 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
897 | setup_btree_index(i, t); | |
898 | } | |
899 | ||
900 | return 0; | |
901 | } | |
902 | ||
903 | /* | |
904 | * Builds the btree to index the map. | |
905 | */ | |
26803b9f | 906 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
907 | { |
908 | int r = 0; | |
909 | unsigned int leaf_nodes; | |
910 | ||
1da177e4 LT |
911 | /* how many indexes will the btree have ? */ |
912 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
913 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
914 | ||
915 | /* leaf layer has already been set up */ | |
916 | t->counts[t->depth - 1] = leaf_nodes; | |
917 | t->index[t->depth - 1] = t->highs; | |
918 | ||
919 | if (t->depth >= 2) | |
920 | r = setup_indexes(t); | |
921 | ||
922 | return r; | |
923 | } | |
924 | ||
26803b9f WD |
925 | /* |
926 | * Register the mapped device for blk_integrity support if | |
927 | * the underlying devices support it. | |
928 | */ | |
929 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
930 | { | |
931 | struct list_head *devices = dm_table_get_devices(t); | |
932 | struct dm_dev_internal *dd; | |
933 | ||
934 | list_for_each_entry(dd, devices, list) | |
935 | if (bdev_get_integrity(dd->dm_dev.bdev)) | |
936 | return blk_integrity_register(dm_disk(md), NULL); | |
937 | ||
938 | return 0; | |
939 | } | |
940 | ||
941 | /* | |
942 | * Prepares the table for use by building the indices, | |
943 | * setting the type, and allocating mempools. | |
944 | */ | |
945 | int dm_table_complete(struct dm_table *t) | |
946 | { | |
947 | int r; | |
948 | ||
949 | r = dm_table_set_type(t); | |
950 | if (r) { | |
951 | DMERR("unable to set table type"); | |
952 | return r; | |
953 | } | |
954 | ||
955 | r = dm_table_build_index(t); | |
956 | if (r) { | |
957 | DMERR("unable to build btrees"); | |
958 | return r; | |
959 | } | |
960 | ||
961 | r = dm_table_prealloc_integrity(t, t->md); | |
962 | if (r) { | |
963 | DMERR("could not register integrity profile."); | |
964 | return r; | |
965 | } | |
966 | ||
967 | r = dm_table_alloc_md_mempools(t); | |
968 | if (r) | |
969 | DMERR("unable to allocate mempools"); | |
970 | ||
971 | return r; | |
972 | } | |
973 | ||
48c9c27b | 974 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
975 | void dm_table_event_callback(struct dm_table *t, |
976 | void (*fn)(void *), void *context) | |
977 | { | |
48c9c27b | 978 | mutex_lock(&_event_lock); |
1da177e4 LT |
979 | t->event_fn = fn; |
980 | t->event_context = context; | |
48c9c27b | 981 | mutex_unlock(&_event_lock); |
1da177e4 LT |
982 | } |
983 | ||
984 | void dm_table_event(struct dm_table *t) | |
985 | { | |
986 | /* | |
987 | * You can no longer call dm_table_event() from interrupt | |
988 | * context, use a bottom half instead. | |
989 | */ | |
990 | BUG_ON(in_interrupt()); | |
991 | ||
48c9c27b | 992 | mutex_lock(&_event_lock); |
1da177e4 LT |
993 | if (t->event_fn) |
994 | t->event_fn(t->event_context); | |
48c9c27b | 995 | mutex_unlock(&_event_lock); |
1da177e4 LT |
996 | } |
997 | ||
998 | sector_t dm_table_get_size(struct dm_table *t) | |
999 | { | |
1000 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1001 | } | |
1002 | ||
1003 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1004 | { | |
14353539 | 1005 | if (index >= t->num_targets) |
1da177e4 LT |
1006 | return NULL; |
1007 | ||
1008 | return t->targets + index; | |
1009 | } | |
1010 | ||
1011 | /* | |
1012 | * Search the btree for the correct target. | |
512875bd JN |
1013 | * |
1014 | * Caller should check returned pointer with dm_target_is_valid() | |
1015 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1016 | */ |
1017 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1018 | { | |
1019 | unsigned int l, n = 0, k = 0; | |
1020 | sector_t *node; | |
1021 | ||
1022 | for (l = 0; l < t->depth; l++) { | |
1023 | n = get_child(n, k); | |
1024 | node = get_node(t, l, n); | |
1025 | ||
1026 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1027 | if (node[k] >= sector) | |
1028 | break; | |
1029 | } | |
1030 | ||
1031 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1032 | } | |
1033 | ||
754c5fc7 MS |
1034 | /* |
1035 | * Establish the new table's queue_limits and validate them. | |
1036 | */ | |
1037 | int dm_calculate_queue_limits(struct dm_table *table, | |
1038 | struct queue_limits *limits) | |
1039 | { | |
1040 | struct dm_target *uninitialized_var(ti); | |
1041 | struct queue_limits ti_limits; | |
1042 | unsigned i = 0; | |
1043 | ||
1044 | blk_set_default_limits(limits); | |
1045 | ||
1046 | while (i < dm_table_get_num_targets(table)) { | |
1047 | blk_set_default_limits(&ti_limits); | |
1048 | ||
1049 | ti = dm_table_get_target(table, i++); | |
1050 | ||
1051 | if (!ti->type->iterate_devices) | |
1052 | goto combine_limits; | |
1053 | ||
1054 | /* | |
1055 | * Combine queue limits of all the devices this target uses. | |
1056 | */ | |
1057 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1058 | &ti_limits); | |
1059 | ||
40bea431 MS |
1060 | /* Set I/O hints portion of queue limits */ |
1061 | if (ti->type->io_hints) | |
1062 | ti->type->io_hints(ti, &ti_limits); | |
1063 | ||
754c5fc7 MS |
1064 | /* |
1065 | * Check each device area is consistent with the target's | |
1066 | * overall queue limits. | |
1067 | */ | |
f6a1ed10 MP |
1068 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1069 | &ti_limits)) | |
754c5fc7 MS |
1070 | return -EINVAL; |
1071 | ||
1072 | combine_limits: | |
1073 | /* | |
1074 | * Merge this target's queue limits into the overall limits | |
1075 | * for the table. | |
1076 | */ | |
1077 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1078 | DMWARN("%s: adding target device " |
754c5fc7 | 1079 | "(start sect %llu len %llu) " |
b27d7f16 | 1080 | "caused an alignment inconsistency", |
754c5fc7 MS |
1081 | dm_device_name(table->md), |
1082 | (unsigned long long) ti->begin, | |
1083 | (unsigned long long) ti->len); | |
1084 | } | |
1085 | ||
1086 | return validate_hardware_logical_block_alignment(table, limits); | |
1087 | } | |
1088 | ||
9c47008d MP |
1089 | /* |
1090 | * Set the integrity profile for this device if all devices used have | |
1091 | * matching profiles. | |
1092 | */ | |
1093 | static void dm_table_set_integrity(struct dm_table *t) | |
1094 | { | |
1095 | struct list_head *devices = dm_table_get_devices(t); | |
1096 | struct dm_dev_internal *prev = NULL, *dd = NULL; | |
1097 | ||
1098 | if (!blk_get_integrity(dm_disk(t->md))) | |
1099 | return; | |
1100 | ||
1101 | list_for_each_entry(dd, devices, list) { | |
1102 | if (prev && | |
1103 | blk_integrity_compare(prev->dm_dev.bdev->bd_disk, | |
1104 | dd->dm_dev.bdev->bd_disk) < 0) { | |
1105 | DMWARN("%s: integrity not set: %s and %s mismatch", | |
1106 | dm_device_name(t->md), | |
1107 | prev->dm_dev.bdev->bd_disk->disk_name, | |
1108 | dd->dm_dev.bdev->bd_disk->disk_name); | |
1109 | goto no_integrity; | |
1110 | } | |
1111 | prev = dd; | |
1112 | } | |
1113 | ||
1114 | if (!prev || !bdev_get_integrity(prev->dm_dev.bdev)) | |
1115 | goto no_integrity; | |
1116 | ||
1117 | blk_integrity_register(dm_disk(t->md), | |
1118 | bdev_get_integrity(prev->dm_dev.bdev)); | |
1119 | ||
1120 | return; | |
1121 | ||
1122 | no_integrity: | |
1123 | blk_integrity_register(dm_disk(t->md), NULL); | |
1124 | ||
1125 | return; | |
1126 | } | |
1127 | ||
754c5fc7 MS |
1128 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1129 | struct queue_limits *limits) | |
1da177e4 LT |
1130 | { |
1131 | /* | |
1197764e | 1132 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1133 | */ |
754c5fc7 | 1134 | q->limits = *limits; |
c9a3f6d6 | 1135 | |
754c5fc7 | 1136 | if (limits->no_cluster) |
c9a3f6d6 | 1137 | queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 1138 | else |
c9a3f6d6 | 1139 | queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 1140 | |
5ae89a87 MS |
1141 | if (!dm_table_supports_discards(t)) |
1142 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1143 | else | |
1144 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1145 | ||
9c47008d | 1146 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1147 | |
1148 | /* | |
1149 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1150 | * visible to other CPUs because, once the flag is set, incoming bios | |
1151 | * are processed by request-based dm, which refers to the queue | |
1152 | * settings. | |
1153 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1154 | * md->deferred where queue settings are not needed yet. | |
1155 | * Those bios are passed to request-based dm at the resume time. | |
1156 | */ | |
1157 | smp_mb(); | |
1158 | if (dm_table_request_based(t)) | |
1159 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1160 | } |
1161 | ||
1162 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1163 | { | |
1164 | return t->num_targets; | |
1165 | } | |
1166 | ||
1167 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1168 | { | |
1169 | return &t->devices; | |
1170 | } | |
1171 | ||
aeb5d727 | 1172 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1173 | { |
1174 | return t->mode; | |
1175 | } | |
1176 | ||
1177 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1178 | { | |
1179 | int i = t->num_targets; | |
1180 | struct dm_target *ti = t->targets; | |
1181 | ||
1182 | while (i--) { | |
1183 | if (postsuspend) { | |
1184 | if (ti->type->postsuspend) | |
1185 | ti->type->postsuspend(ti); | |
1186 | } else if (ti->type->presuspend) | |
1187 | ti->type->presuspend(ti); | |
1188 | ||
1189 | ti++; | |
1190 | } | |
1191 | } | |
1192 | ||
1193 | void dm_table_presuspend_targets(struct dm_table *t) | |
1194 | { | |
cf222b37 AK |
1195 | if (!t) |
1196 | return; | |
1197 | ||
e8488d08 | 1198 | suspend_targets(t, 0); |
1da177e4 LT |
1199 | } |
1200 | ||
1201 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1202 | { | |
cf222b37 AK |
1203 | if (!t) |
1204 | return; | |
1205 | ||
e8488d08 | 1206 | suspend_targets(t, 1); |
1da177e4 LT |
1207 | } |
1208 | ||
8757b776 | 1209 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1210 | { |
8757b776 MB |
1211 | int i, r = 0; |
1212 | ||
1213 | for (i = 0; i < t->num_targets; i++) { | |
1214 | struct dm_target *ti = t->targets + i; | |
1215 | ||
1216 | if (!ti->type->preresume) | |
1217 | continue; | |
1218 | ||
1219 | r = ti->type->preresume(ti); | |
1220 | if (r) | |
1221 | return r; | |
1222 | } | |
1da177e4 LT |
1223 | |
1224 | for (i = 0; i < t->num_targets; i++) { | |
1225 | struct dm_target *ti = t->targets + i; | |
1226 | ||
1227 | if (ti->type->resume) | |
1228 | ti->type->resume(ti); | |
1229 | } | |
8757b776 MB |
1230 | |
1231 | return 0; | |
1da177e4 LT |
1232 | } |
1233 | ||
1234 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | |
1235 | { | |
82b1519b | 1236 | struct dm_dev_internal *dd; |
afb24528 | 1237 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 LT |
1238 | int r = 0; |
1239 | ||
afb24528 | 1240 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1241 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1242 | char b[BDEVNAME_SIZE]; |
1243 | ||
1244 | if (likely(q)) | |
1245 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1246 | else | |
1247 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1248 | dm_device_name(t->md), | |
1249 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1250 | } |
1251 | ||
1252 | return r; | |
1253 | } | |
1254 | ||
cec47e3d KU |
1255 | int dm_table_any_busy_target(struct dm_table *t) |
1256 | { | |
1257 | unsigned i; | |
1258 | struct dm_target *ti; | |
1259 | ||
1260 | for (i = 0; i < t->num_targets; i++) { | |
1261 | ti = t->targets + i; | |
1262 | if (ti->type->busy && ti->type->busy(ti)) | |
1263 | return 1; | |
1264 | } | |
1265 | ||
1266 | return 0; | |
1267 | } | |
1268 | ||
1da177e4 LT |
1269 | void dm_table_unplug_all(struct dm_table *t) |
1270 | { | |
82b1519b | 1271 | struct dm_dev_internal *dd; |
afb24528 | 1272 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 | 1273 | |
afb24528 | 1274 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1275 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1276 | char b[BDEVNAME_SIZE]; |
1277 | ||
1278 | if (likely(q)) | |
1279 | blk_unplug(q); | |
1280 | else | |
1281 | DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s", | |
1282 | dm_device_name(t->md), | |
1283 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1284 | } |
1285 | } | |
1286 | ||
1134e5ae MA |
1287 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1288 | { | |
1134e5ae MA |
1289 | return t->md; |
1290 | } | |
1291 | ||
5ae89a87 MS |
1292 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1293 | sector_t start, sector_t len, void *data) | |
1294 | { | |
1295 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1296 | ||
1297 | return q && blk_queue_discard(q); | |
1298 | } | |
1299 | ||
1300 | bool dm_table_supports_discards(struct dm_table *t) | |
1301 | { | |
1302 | struct dm_target *ti; | |
1303 | unsigned i = 0; | |
1304 | ||
1305 | if (!t->discards_supported) | |
1306 | return 0; | |
1307 | ||
1308 | /* | |
1309 | * Ensure that at least one underlying device supports discards. | |
1310 | * t->devices includes internal dm devices such as mirror logs | |
1311 | * so we need to use iterate_devices here, which targets | |
1312 | * supporting discard must provide. | |
1313 | */ | |
1314 | while (i < dm_table_get_num_targets(t)) { | |
1315 | ti = dm_table_get_target(t, i++); | |
1316 | ||
1317 | if (ti->type->iterate_devices && | |
1318 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1319 | return 1; | |
1320 | } | |
1321 | ||
1322 | return 0; | |
1323 | } | |
1324 | ||
1da177e4 LT |
1325 | EXPORT_SYMBOL(dm_vcalloc); |
1326 | EXPORT_SYMBOL(dm_get_device); | |
1327 | EXPORT_SYMBOL(dm_put_device); | |
1328 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1329 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1330 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1331 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1332 | EXPORT_SYMBOL(dm_table_put); |
1333 | EXPORT_SYMBOL(dm_table_get); | |
1334 | EXPORT_SYMBOL(dm_table_unplug_all); |