]>
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 | 57 | unsigned discards_supported:1; |
a91a2785 | 58 | unsigned integrity_supported:1; |
5ae89a87 | 59 | |
1da177e4 LT |
60 | /* |
61 | * Indicates the rw permissions for the new logical | |
62 | * device. This should be a combination of FMODE_READ | |
63 | * and FMODE_WRITE. | |
64 | */ | |
aeb5d727 | 65 | fmode_t mode; |
1da177e4 LT |
66 | |
67 | /* a list of devices used by this table */ | |
68 | struct list_head devices; | |
69 | ||
1da177e4 LT |
70 | /* events get handed up using this callback */ |
71 | void (*event_fn)(void *); | |
72 | void *event_context; | |
e6ee8c0b KU |
73 | |
74 | struct dm_md_mempools *mempools; | |
9d357b07 N |
75 | |
76 | struct list_head target_callbacks; | |
1da177e4 LT |
77 | }; |
78 | ||
79 | /* | |
80 | * Similar to ceiling(log_size(n)) | |
81 | */ | |
82 | static unsigned int int_log(unsigned int n, unsigned int base) | |
83 | { | |
84 | int result = 0; | |
85 | ||
86 | while (n > 1) { | |
87 | n = dm_div_up(n, base); | |
88 | result++; | |
89 | } | |
90 | ||
91 | return result; | |
92 | } | |
93 | ||
1da177e4 LT |
94 | /* |
95 | * Calculate the index of the child node of the n'th node k'th key. | |
96 | */ | |
97 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
98 | { | |
99 | return (n * CHILDREN_PER_NODE) + k; | |
100 | } | |
101 | ||
102 | /* | |
103 | * Return the n'th node of level l from table t. | |
104 | */ | |
105 | static inline sector_t *get_node(struct dm_table *t, | |
106 | unsigned int l, unsigned int n) | |
107 | { | |
108 | return t->index[l] + (n * KEYS_PER_NODE); | |
109 | } | |
110 | ||
111 | /* | |
112 | * Return the highest key that you could lookup from the n'th | |
113 | * node on level l of the btree. | |
114 | */ | |
115 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
116 | { | |
117 | for (; l < t->depth - 1; l++) | |
118 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
119 | ||
120 | if (n >= t->counts[l]) | |
121 | return (sector_t) - 1; | |
122 | ||
123 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
124 | } | |
125 | ||
126 | /* | |
127 | * Fills in a level of the btree based on the highs of the level | |
128 | * below it. | |
129 | */ | |
130 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
131 | { | |
132 | unsigned int n, k; | |
133 | sector_t *node; | |
134 | ||
135 | for (n = 0U; n < t->counts[l]; n++) { | |
136 | node = get_node(t, l, n); | |
137 | ||
138 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
139 | node[k] = high(t, l + 1, get_child(n, k)); | |
140 | } | |
141 | ||
142 | return 0; | |
143 | } | |
144 | ||
145 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
146 | { | |
147 | unsigned long size; | |
148 | void *addr; | |
149 | ||
150 | /* | |
151 | * Check that we're not going to overflow. | |
152 | */ | |
153 | if (nmemb > (ULONG_MAX / elem_size)) | |
154 | return NULL; | |
155 | ||
156 | size = nmemb * elem_size; | |
157 | addr = vmalloc(size); | |
158 | if (addr) | |
159 | memset(addr, 0, size); | |
160 | ||
161 | return addr; | |
162 | } | |
163 | ||
164 | /* | |
165 | * highs, and targets are managed as dynamic arrays during a | |
166 | * table load. | |
167 | */ | |
168 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
169 | { | |
170 | sector_t *n_highs; | |
171 | struct dm_target *n_targets; | |
172 | int n = t->num_targets; | |
173 | ||
174 | /* | |
175 | * Allocate both the target array and offset array at once. | |
512875bd JN |
176 | * Append an empty entry to catch sectors beyond the end of |
177 | * the device. | |
1da177e4 | 178 | */ |
512875bd | 179 | n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) + |
1da177e4 LT |
180 | sizeof(sector_t)); |
181 | if (!n_highs) | |
182 | return -ENOMEM; | |
183 | ||
184 | n_targets = (struct dm_target *) (n_highs + num); | |
185 | ||
186 | if (n) { | |
187 | memcpy(n_highs, t->highs, sizeof(*n_highs) * n); | |
188 | memcpy(n_targets, t->targets, sizeof(*n_targets) * n); | |
189 | } | |
190 | ||
191 | memset(n_highs + n, -1, sizeof(*n_highs) * (num - n)); | |
192 | vfree(t->highs); | |
193 | ||
194 | t->num_allocated = num; | |
195 | t->highs = n_highs; | |
196 | t->targets = n_targets; | |
197 | ||
198 | return 0; | |
199 | } | |
200 | ||
aeb5d727 | 201 | int dm_table_create(struct dm_table **result, fmode_t mode, |
1134e5ae | 202 | unsigned num_targets, struct mapped_device *md) |
1da177e4 | 203 | { |
094262db | 204 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
205 | |
206 | if (!t) | |
207 | return -ENOMEM; | |
208 | ||
1da177e4 | 209 | INIT_LIST_HEAD(&t->devices); |
9d357b07 | 210 | INIT_LIST_HEAD(&t->target_callbacks); |
d5816876 | 211 | atomic_set(&t->holders, 0); |
5ae89a87 | 212 | t->discards_supported = 1; |
1da177e4 LT |
213 | |
214 | if (!num_targets) | |
215 | num_targets = KEYS_PER_NODE; | |
216 | ||
217 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
218 | ||
219 | if (alloc_targets(t, num_targets)) { | |
220 | kfree(t); | |
221 | t = NULL; | |
222 | return -ENOMEM; | |
223 | } | |
224 | ||
225 | t->mode = mode; | |
1134e5ae | 226 | t->md = md; |
1da177e4 LT |
227 | *result = t; |
228 | return 0; | |
229 | } | |
230 | ||
231 | static void free_devices(struct list_head *devices) | |
232 | { | |
233 | struct list_head *tmp, *next; | |
234 | ||
afb24528 | 235 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
236 | struct dm_dev_internal *dd = |
237 | list_entry(tmp, struct dm_dev_internal, list); | |
1b6da754 JB |
238 | DMWARN("dm_table_destroy: dm_put_device call missing for %s", |
239 | dd->dm_dev.name); | |
1da177e4 LT |
240 | kfree(dd); |
241 | } | |
242 | } | |
243 | ||
d5816876 | 244 | void dm_table_destroy(struct dm_table *t) |
1da177e4 LT |
245 | { |
246 | unsigned int i; | |
247 | ||
a7940155 AK |
248 | if (!t) |
249 | return; | |
250 | ||
d5816876 MP |
251 | while (atomic_read(&t->holders)) |
252 | msleep(1); | |
253 | smp_mb(); | |
254 | ||
26803b9f | 255 | /* free the indexes */ |
1da177e4 LT |
256 | if (t->depth >= 2) |
257 | vfree(t->index[t->depth - 2]); | |
258 | ||
259 | /* free the targets */ | |
260 | for (i = 0; i < t->num_targets; i++) { | |
261 | struct dm_target *tgt = t->targets + i; | |
262 | ||
263 | if (tgt->type->dtr) | |
264 | tgt->type->dtr(tgt); | |
265 | ||
266 | dm_put_target_type(tgt->type); | |
267 | } | |
268 | ||
269 | vfree(t->highs); | |
270 | ||
271 | /* free the device list */ | |
1b6da754 | 272 | if (t->devices.next != &t->devices) |
1da177e4 | 273 | free_devices(&t->devices); |
1da177e4 | 274 | |
e6ee8c0b KU |
275 | dm_free_md_mempools(t->mempools); |
276 | ||
1da177e4 LT |
277 | kfree(t); |
278 | } | |
279 | ||
280 | void dm_table_get(struct dm_table *t) | |
281 | { | |
282 | atomic_inc(&t->holders); | |
283 | } | |
284 | ||
285 | void dm_table_put(struct dm_table *t) | |
286 | { | |
287 | if (!t) | |
288 | return; | |
289 | ||
d5816876 MP |
290 | smp_mb__before_atomic_dec(); |
291 | atomic_dec(&t->holders); | |
1da177e4 LT |
292 | } |
293 | ||
294 | /* | |
295 | * Checks to see if we need to extend highs or targets. | |
296 | */ | |
297 | static inline int check_space(struct dm_table *t) | |
298 | { | |
299 | if (t->num_targets >= t->num_allocated) | |
300 | return alloc_targets(t, t->num_allocated * 2); | |
301 | ||
302 | return 0; | |
303 | } | |
304 | ||
1da177e4 LT |
305 | /* |
306 | * See if we've already got a device in the list. | |
307 | */ | |
82b1519b | 308 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 309 | { |
82b1519b | 310 | struct dm_dev_internal *dd; |
1da177e4 LT |
311 | |
312 | list_for_each_entry (dd, l, list) | |
82b1519b | 313 | if (dd->dm_dev.bdev->bd_dev == dev) |
1da177e4 LT |
314 | return dd; |
315 | ||
316 | return NULL; | |
317 | } | |
318 | ||
319 | /* | |
320 | * Open a device so we can use it as a map destination. | |
321 | */ | |
82b1519b MP |
322 | static int open_dev(struct dm_dev_internal *d, dev_t dev, |
323 | struct mapped_device *md) | |
1da177e4 LT |
324 | { |
325 | static char *_claim_ptr = "I belong to device-mapper"; | |
326 | struct block_device *bdev; | |
327 | ||
328 | int r; | |
329 | ||
82b1519b | 330 | BUG_ON(d->dm_dev.bdev); |
1da177e4 | 331 | |
d4d77629 | 332 | bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr); |
1da177e4 LT |
333 | if (IS_ERR(bdev)) |
334 | return PTR_ERR(bdev); | |
e09b457b | 335 | |
e09b457b TH |
336 | r = bd_link_disk_holder(bdev, dm_disk(md)); |
337 | if (r) { | |
e525fd89 | 338 | blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL); |
e09b457b TH |
339 | return r; |
340 | } | |
341 | ||
342 | d->dm_dev.bdev = bdev; | |
343 | return 0; | |
1da177e4 LT |
344 | } |
345 | ||
346 | /* | |
347 | * Close a device that we've been using. | |
348 | */ | |
82b1519b | 349 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 350 | { |
82b1519b | 351 | if (!d->dm_dev.bdev) |
1da177e4 LT |
352 | return; |
353 | ||
49731baa | 354 | bd_unlink_disk_holder(d->dm_dev.bdev, dm_disk(md)); |
e525fd89 | 355 | blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL); |
82b1519b | 356 | d->dm_dev.bdev = NULL; |
1da177e4 LT |
357 | } |
358 | ||
359 | /* | |
f6a1ed10 | 360 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 361 | */ |
f6a1ed10 MP |
362 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
363 | sector_t start, sector_t len, void *data) | |
1da177e4 | 364 | { |
754c5fc7 MS |
365 | struct queue_limits *limits = data; |
366 | struct block_device *bdev = dev->bdev; | |
367 | sector_t dev_size = | |
368 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 369 | unsigned short logical_block_size_sectors = |
754c5fc7 | 370 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 371 | char b[BDEVNAME_SIZE]; |
2cd54d9b MA |
372 | |
373 | if (!dev_size) | |
f6a1ed10 | 374 | return 0; |
2cd54d9b | 375 | |
5dea271b | 376 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
377 | DMWARN("%s: %s too small for target: " |
378 | "start=%llu, len=%llu, dev_size=%llu", | |
379 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
380 | (unsigned long long)start, | |
381 | (unsigned long long)len, | |
382 | (unsigned long long)dev_size); | |
f6a1ed10 | 383 | return 1; |
02acc3a4 MS |
384 | } |
385 | ||
386 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 387 | return 0; |
02acc3a4 MS |
388 | |
389 | if (start & (logical_block_size_sectors - 1)) { | |
390 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 391 | "logical block size %u of %s", |
02acc3a4 MS |
392 | dm_device_name(ti->table->md), |
393 | (unsigned long long)start, | |
754c5fc7 | 394 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 395 | return 1; |
02acc3a4 MS |
396 | } |
397 | ||
5dea271b | 398 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 399 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 400 | "logical block size %u of %s", |
02acc3a4 | 401 | dm_device_name(ti->table->md), |
5dea271b | 402 | (unsigned long long)len, |
754c5fc7 | 403 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 404 | return 1; |
02acc3a4 MS |
405 | } |
406 | ||
f6a1ed10 | 407 | return 0; |
1da177e4 LT |
408 | } |
409 | ||
410 | /* | |
570b9d96 | 411 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 412 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
413 | * device and not to touch the existing bdev field in case |
414 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 415 | */ |
aeb5d727 | 416 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 417 | struct mapped_device *md) |
1da177e4 LT |
418 | { |
419 | int r; | |
570b9d96 | 420 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 421 | |
570b9d96 AK |
422 | dd_new = dd_old = *dd; |
423 | ||
424 | dd_new.dm_dev.mode |= new_mode; | |
425 | dd_new.dm_dev.bdev = NULL; | |
426 | ||
427 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
428 | if (r) | |
429 | return r; | |
1da177e4 | 430 | |
82b1519b | 431 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 432 | close_dev(&dd_old, md); |
1da177e4 | 433 | |
570b9d96 | 434 | return 0; |
1da177e4 LT |
435 | } |
436 | ||
437 | /* | |
438 | * Add a device to the list, or just increment the usage count if | |
439 | * it's already present. | |
440 | */ | |
441 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
8215d6ec | 442 | const char *path, fmode_t mode, struct dm_dev **result) |
1da177e4 LT |
443 | { |
444 | int r; | |
69a2ce72 | 445 | dev_t uninitialized_var(dev); |
82b1519b | 446 | struct dm_dev_internal *dd; |
1da177e4 LT |
447 | unsigned int major, minor; |
448 | ||
547bc926 | 449 | BUG_ON(!t); |
1da177e4 LT |
450 | |
451 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
452 | /* Extract the major/minor numbers */ | |
453 | dev = MKDEV(major, minor); | |
454 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
455 | return -EOVERFLOW; | |
456 | } else { | |
457 | /* convert the path to a device */ | |
72e8264e CH |
458 | struct block_device *bdev = lookup_bdev(path); |
459 | ||
460 | if (IS_ERR(bdev)) | |
461 | return PTR_ERR(bdev); | |
462 | dev = bdev->bd_dev; | |
463 | bdput(bdev); | |
1da177e4 LT |
464 | } |
465 | ||
466 | dd = find_device(&t->devices, dev); | |
467 | if (!dd) { | |
468 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
469 | if (!dd) | |
470 | return -ENOMEM; | |
471 | ||
82b1519b MP |
472 | dd->dm_dev.mode = mode; |
473 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 474 | |
f165921d | 475 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
476 | kfree(dd); |
477 | return r; | |
478 | } | |
479 | ||
82b1519b | 480 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
481 | |
482 | atomic_set(&dd->count, 0); | |
483 | list_add(&dd->list, &t->devices); | |
484 | ||
82b1519b | 485 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 486 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
487 | if (r) |
488 | return r; | |
489 | } | |
490 | atomic_inc(&dd->count); | |
491 | ||
82b1519b | 492 | *result = &dd->dm_dev; |
1da177e4 LT |
493 | return 0; |
494 | } | |
495 | ||
754c5fc7 | 496 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 497 | sector_t start, sector_t len, void *data) |
1da177e4 | 498 | { |
754c5fc7 MS |
499 | struct queue_limits *limits = data; |
500 | struct block_device *bdev = dev->bdev; | |
165125e1 | 501 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
502 | char b[BDEVNAME_SIZE]; |
503 | ||
504 | if (unlikely(!q)) { | |
505 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
506 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 507 | return 0; |
0c2322e4 | 508 | } |
3cb40214 | 509 | |
b27d7f16 MP |
510 | if (bdev_stack_limits(limits, bdev, start) < 0) |
511 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
512 | "physical_block_size=%u, logical_block_size=%u, " |
513 | "alignment_offset=%u, start=%llu", | |
514 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
515 | q->limits.physical_block_size, | |
516 | q->limits.logical_block_size, | |
517 | q->limits.alignment_offset, | |
b27d7f16 | 518 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 519 | |
9980c638 MB |
520 | /* |
521 | * Check if merge fn is supported. | |
522 | * If not we'll force DM to use PAGE_SIZE or | |
523 | * smaller I/O, just to be safe. | |
3cb40214 | 524 | */ |
9980c638 MB |
525 | |
526 | if (q->merge_bvec_fn && !ti->type->merge) | |
72d4cd9f MS |
527 | blk_limits_max_hw_sectors(limits, |
528 | (unsigned int) (PAGE_SIZE >> 9)); | |
754c5fc7 | 529 | return 0; |
3cb40214 BR |
530 | } |
531 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 532 | |
8215d6ec NK |
533 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
534 | struct dm_dev **result) | |
3cb40214 | 535 | { |
8215d6ec | 536 | return __table_get_device(ti->table, ti, path, mode, result); |
1da177e4 LT |
537 | } |
538 | ||
754c5fc7 | 539 | |
1da177e4 LT |
540 | /* |
541 | * Decrement a devices use count and remove it if necessary. | |
542 | */ | |
82b1519b | 543 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 544 | { |
82b1519b MP |
545 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
546 | dm_dev); | |
547 | ||
1da177e4 | 548 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 549 | close_dev(dd, ti->table->md); |
1da177e4 LT |
550 | list_del(&dd->list); |
551 | kfree(dd); | |
552 | } | |
553 | } | |
554 | ||
555 | /* | |
556 | * Checks to see if the target joins onto the end of the table. | |
557 | */ | |
558 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
559 | { | |
560 | struct dm_target *prev; | |
561 | ||
562 | if (!table->num_targets) | |
563 | return !ti->begin; | |
564 | ||
565 | prev = &table->targets[table->num_targets - 1]; | |
566 | return (ti->begin == (prev->begin + prev->len)); | |
567 | } | |
568 | ||
569 | /* | |
570 | * Used to dynamically allocate the arg array. | |
571 | */ | |
572 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
573 | { | |
574 | char **argv; | |
575 | unsigned new_size; | |
576 | ||
577 | new_size = *array_size ? *array_size * 2 : 64; | |
578 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
579 | if (argv) { | |
580 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
581 | *array_size = new_size; | |
582 | } | |
583 | ||
584 | kfree(old_argv); | |
585 | return argv; | |
586 | } | |
587 | ||
588 | /* | |
589 | * Destructively splits up the argument list to pass to ctr. | |
590 | */ | |
591 | int dm_split_args(int *argc, char ***argvp, char *input) | |
592 | { | |
593 | char *start, *end = input, *out, **argv = NULL; | |
594 | unsigned array_size = 0; | |
595 | ||
596 | *argc = 0; | |
814d6862 DT |
597 | |
598 | if (!input) { | |
599 | *argvp = NULL; | |
600 | return 0; | |
601 | } | |
602 | ||
1da177e4 LT |
603 | argv = realloc_argv(&array_size, argv); |
604 | if (!argv) | |
605 | return -ENOMEM; | |
606 | ||
607 | while (1) { | |
1da177e4 | 608 | /* Skip whitespace */ |
e7d2860b | 609 | start = skip_spaces(end); |
1da177e4 LT |
610 | |
611 | if (!*start) | |
612 | break; /* success, we hit the end */ | |
613 | ||
614 | /* 'out' is used to remove any back-quotes */ | |
615 | end = out = start; | |
616 | while (*end) { | |
617 | /* Everything apart from '\0' can be quoted */ | |
618 | if (*end == '\\' && *(end + 1)) { | |
619 | *out++ = *(end + 1); | |
620 | end += 2; | |
621 | continue; | |
622 | } | |
623 | ||
624 | if (isspace(*end)) | |
625 | break; /* end of token */ | |
626 | ||
627 | *out++ = *end++; | |
628 | } | |
629 | ||
630 | /* have we already filled the array ? */ | |
631 | if ((*argc + 1) > array_size) { | |
632 | argv = realloc_argv(&array_size, argv); | |
633 | if (!argv) | |
634 | return -ENOMEM; | |
635 | } | |
636 | ||
637 | /* we know this is whitespace */ | |
638 | if (*end) | |
639 | end++; | |
640 | ||
641 | /* terminate the string and put it in the array */ | |
642 | *out = '\0'; | |
643 | argv[*argc] = start; | |
644 | (*argc)++; | |
645 | } | |
646 | ||
647 | *argvp = argv; | |
648 | return 0; | |
649 | } | |
650 | ||
be6d4305 MS |
651 | /* |
652 | * Impose necessary and sufficient conditions on a devices's table such | |
653 | * that any incoming bio which respects its logical_block_size can be | |
654 | * processed successfully. If it falls across the boundary between | |
655 | * two or more targets, the size of each piece it gets split into must | |
656 | * be compatible with the logical_block_size of the target processing it. | |
657 | */ | |
754c5fc7 MS |
658 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
659 | struct queue_limits *limits) | |
be6d4305 MS |
660 | { |
661 | /* | |
662 | * This function uses arithmetic modulo the logical_block_size | |
663 | * (in units of 512-byte sectors). | |
664 | */ | |
665 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 666 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
667 | |
668 | /* | |
669 | * Offset of the start of the next table entry, mod logical_block_size. | |
670 | */ | |
671 | unsigned short next_target_start = 0; | |
672 | ||
673 | /* | |
674 | * Given an aligned bio that extends beyond the end of a | |
675 | * target, how many sectors must the next target handle? | |
676 | */ | |
677 | unsigned short remaining = 0; | |
678 | ||
679 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 680 | struct queue_limits ti_limits; |
be6d4305 MS |
681 | unsigned i = 0; |
682 | ||
683 | /* | |
684 | * Check each entry in the table in turn. | |
685 | */ | |
686 | while (i < dm_table_get_num_targets(table)) { | |
687 | ti = dm_table_get_target(table, i++); | |
688 | ||
754c5fc7 MS |
689 | blk_set_default_limits(&ti_limits); |
690 | ||
691 | /* combine all target devices' limits */ | |
692 | if (ti->type->iterate_devices) | |
693 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
694 | &ti_limits); | |
695 | ||
be6d4305 MS |
696 | /* |
697 | * If the remaining sectors fall entirely within this | |
698 | * table entry are they compatible with its logical_block_size? | |
699 | */ | |
700 | if (remaining < ti->len && | |
754c5fc7 | 701 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
702 | SECTOR_SHIFT) - 1)) |
703 | break; /* Error */ | |
704 | ||
705 | next_target_start = | |
706 | (unsigned short) ((next_target_start + ti->len) & | |
707 | (device_logical_block_size_sects - 1)); | |
708 | remaining = next_target_start ? | |
709 | device_logical_block_size_sects - next_target_start : 0; | |
710 | } | |
711 | ||
712 | if (remaining) { | |
713 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 714 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
715 | dm_device_name(table->md), i, |
716 | (unsigned long long) ti->begin, | |
717 | (unsigned long long) ti->len, | |
754c5fc7 | 718 | limits->logical_block_size); |
be6d4305 MS |
719 | return -EINVAL; |
720 | } | |
721 | ||
722 | return 0; | |
723 | } | |
724 | ||
1da177e4 LT |
725 | int dm_table_add_target(struct dm_table *t, const char *type, |
726 | sector_t start, sector_t len, char *params) | |
727 | { | |
728 | int r = -EINVAL, argc; | |
729 | char **argv; | |
730 | struct dm_target *tgt; | |
731 | ||
732 | if ((r = check_space(t))) | |
733 | return r; | |
734 | ||
735 | tgt = t->targets + t->num_targets; | |
736 | memset(tgt, 0, sizeof(*tgt)); | |
737 | ||
738 | if (!len) { | |
72d94861 | 739 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
740 | return -EINVAL; |
741 | } | |
742 | ||
743 | tgt->type = dm_get_target_type(type); | |
744 | if (!tgt->type) { | |
72d94861 AK |
745 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
746 | type); | |
1da177e4 LT |
747 | return -EINVAL; |
748 | } | |
749 | ||
750 | tgt->table = t; | |
751 | tgt->begin = start; | |
752 | tgt->len = len; | |
753 | tgt->error = "Unknown error"; | |
754 | ||
755 | /* | |
756 | * Does this target adjoin the previous one ? | |
757 | */ | |
758 | if (!adjoin(t, tgt)) { | |
759 | tgt->error = "Gap in table"; | |
760 | r = -EINVAL; | |
761 | goto bad; | |
762 | } | |
763 | ||
764 | r = dm_split_args(&argc, &argv, params); | |
765 | if (r) { | |
766 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
767 | goto bad; | |
768 | } | |
769 | ||
770 | r = tgt->type->ctr(tgt, argc, argv); | |
771 | kfree(argv); | |
772 | if (r) | |
773 | goto bad; | |
774 | ||
775 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
776 | ||
5ae89a87 MS |
777 | if (!tgt->num_discard_requests) |
778 | t->discards_supported = 0; | |
779 | ||
1da177e4 LT |
780 | return 0; |
781 | ||
782 | bad: | |
72d94861 | 783 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
784 | dm_put_target_type(tgt->type); |
785 | return r; | |
786 | } | |
787 | ||
26803b9f | 788 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
789 | { |
790 | unsigned i; | |
791 | unsigned bio_based = 0, request_based = 0; | |
792 | struct dm_target *tgt; | |
793 | struct dm_dev_internal *dd; | |
794 | struct list_head *devices; | |
795 | ||
796 | for (i = 0; i < t->num_targets; i++) { | |
797 | tgt = t->targets + i; | |
798 | if (dm_target_request_based(tgt)) | |
799 | request_based = 1; | |
800 | else | |
801 | bio_based = 1; | |
802 | ||
803 | if (bio_based && request_based) { | |
804 | DMWARN("Inconsistent table: different target types" | |
805 | " can't be mixed up"); | |
806 | return -EINVAL; | |
807 | } | |
808 | } | |
809 | ||
810 | if (bio_based) { | |
811 | /* We must use this table as bio-based */ | |
812 | t->type = DM_TYPE_BIO_BASED; | |
813 | return 0; | |
814 | } | |
815 | ||
816 | BUG_ON(!request_based); /* No targets in this table */ | |
817 | ||
818 | /* Non-request-stackable devices can't be used for request-based dm */ | |
819 | devices = dm_table_get_devices(t); | |
820 | list_for_each_entry(dd, devices, list) { | |
821 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
822 | DMWARN("table load rejected: including" | |
823 | " non-request-stackable devices"); | |
824 | return -EINVAL; | |
825 | } | |
826 | } | |
827 | ||
828 | /* | |
829 | * Request-based dm supports only tables that have a single target now. | |
830 | * To support multiple targets, request splitting support is needed, | |
831 | * and that needs lots of changes in the block-layer. | |
832 | * (e.g. request completion process for partial completion.) | |
833 | */ | |
834 | if (t->num_targets > 1) { | |
835 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
836 | return -EINVAL; | |
837 | } | |
838 | ||
839 | t->type = DM_TYPE_REQUEST_BASED; | |
840 | ||
841 | return 0; | |
842 | } | |
843 | ||
844 | unsigned dm_table_get_type(struct dm_table *t) | |
845 | { | |
846 | return t->type; | |
847 | } | |
848 | ||
849 | bool dm_table_request_based(struct dm_table *t) | |
850 | { | |
851 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
852 | } | |
853 | ||
854 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
855 | { | |
856 | unsigned type = dm_table_get_type(t); | |
857 | ||
858 | if (unlikely(type == DM_TYPE_NONE)) { | |
859 | DMWARN("no table type is set, can't allocate mempools"); | |
860 | return -EINVAL; | |
861 | } | |
862 | ||
a91a2785 | 863 | t->mempools = dm_alloc_md_mempools(type, t->integrity_supported); |
e6ee8c0b KU |
864 | if (!t->mempools) |
865 | return -ENOMEM; | |
866 | ||
867 | return 0; | |
868 | } | |
869 | ||
870 | void dm_table_free_md_mempools(struct dm_table *t) | |
871 | { | |
872 | dm_free_md_mempools(t->mempools); | |
873 | t->mempools = NULL; | |
874 | } | |
875 | ||
876 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
877 | { | |
878 | return t->mempools; | |
879 | } | |
880 | ||
1da177e4 LT |
881 | static int setup_indexes(struct dm_table *t) |
882 | { | |
883 | int i; | |
884 | unsigned int total = 0; | |
885 | sector_t *indexes; | |
886 | ||
887 | /* allocate the space for *all* the indexes */ | |
888 | for (i = t->depth - 2; i >= 0; i--) { | |
889 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
890 | total += t->counts[i]; | |
891 | } | |
892 | ||
893 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
894 | if (!indexes) | |
895 | return -ENOMEM; | |
896 | ||
897 | /* set up internal nodes, bottom-up */ | |
82d601dc | 898 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
899 | t->index[i] = indexes; |
900 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
901 | setup_btree_index(i, t); | |
902 | } | |
903 | ||
904 | return 0; | |
905 | } | |
906 | ||
907 | /* | |
908 | * Builds the btree to index the map. | |
909 | */ | |
26803b9f | 910 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
911 | { |
912 | int r = 0; | |
913 | unsigned int leaf_nodes; | |
914 | ||
1da177e4 LT |
915 | /* how many indexes will the btree have ? */ |
916 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
917 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
918 | ||
919 | /* leaf layer has already been set up */ | |
920 | t->counts[t->depth - 1] = leaf_nodes; | |
921 | t->index[t->depth - 1] = t->highs; | |
922 | ||
923 | if (t->depth >= 2) | |
924 | r = setup_indexes(t); | |
925 | ||
926 | return r; | |
927 | } | |
928 | ||
a63a5cf8 MS |
929 | /* |
930 | * Get a disk whose integrity profile reflects the table's profile. | |
931 | * If %match_all is true, all devices' profiles must match. | |
932 | * If %match_all is false, all devices must at least have an | |
933 | * allocated integrity profile; but uninitialized is ok. | |
934 | * Returns NULL if integrity support was inconsistent or unavailable. | |
935 | */ | |
936 | static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t, | |
937 | bool match_all) | |
938 | { | |
939 | struct list_head *devices = dm_table_get_devices(t); | |
940 | struct dm_dev_internal *dd = NULL; | |
941 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
942 | ||
943 | list_for_each_entry(dd, devices, list) { | |
944 | template_disk = dd->dm_dev.bdev->bd_disk; | |
945 | if (!blk_get_integrity(template_disk)) | |
946 | goto no_integrity; | |
947 | if (!match_all && !blk_integrity_is_initialized(template_disk)) | |
948 | continue; /* skip uninitialized profiles */ | |
949 | else if (prev_disk && | |
950 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
951 | goto no_integrity; | |
952 | prev_disk = template_disk; | |
953 | } | |
954 | ||
955 | return template_disk; | |
956 | ||
957 | no_integrity: | |
958 | if (prev_disk) | |
959 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
960 | dm_device_name(t->md), | |
961 | prev_disk->disk_name, | |
962 | template_disk->disk_name); | |
963 | return NULL; | |
964 | } | |
965 | ||
26803b9f WD |
966 | /* |
967 | * Register the mapped device for blk_integrity support if | |
a63a5cf8 MS |
968 | * the underlying devices have an integrity profile. But all devices |
969 | * may not have matching profiles (checking all devices isn't reliable | |
970 | * during table load because this table may use other DM device(s) which | |
971 | * must be resumed before they will have an initialized integity profile). | |
972 | * Stacked DM devices force a 2 stage integrity profile validation: | |
973 | * 1 - during load, validate all initialized integrity profiles match | |
974 | * 2 - during resume, validate all integrity profiles match | |
26803b9f WD |
975 | */ |
976 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
977 | { | |
a63a5cf8 | 978 | struct gendisk *template_disk = NULL; |
26803b9f | 979 | |
a63a5cf8 MS |
980 | template_disk = dm_table_get_integrity_disk(t, false); |
981 | if (!template_disk) | |
982 | return 0; | |
26803b9f | 983 | |
a63a5cf8 MS |
984 | if (!blk_integrity_is_initialized(dm_disk(md))) { |
985 | t->integrity_supported = 1; | |
986 | return blk_integrity_register(dm_disk(md), NULL); | |
987 | } | |
988 | ||
989 | /* | |
990 | * If DM device already has an initalized integrity | |
991 | * profile the new profile should not conflict. | |
992 | */ | |
993 | if (blk_integrity_is_initialized(template_disk) && | |
994 | blk_integrity_compare(dm_disk(md), template_disk) < 0) { | |
995 | DMWARN("%s: conflict with existing integrity profile: " | |
996 | "%s profile mismatch", | |
997 | dm_device_name(t->md), | |
998 | template_disk->disk_name); | |
999 | return 1; | |
1000 | } | |
1001 | ||
1002 | /* Preserve existing initialized integrity profile */ | |
1003 | t->integrity_supported = 1; | |
26803b9f WD |
1004 | return 0; |
1005 | } | |
1006 | ||
1007 | /* | |
1008 | * Prepares the table for use by building the indices, | |
1009 | * setting the type, and allocating mempools. | |
1010 | */ | |
1011 | int dm_table_complete(struct dm_table *t) | |
1012 | { | |
1013 | int r; | |
1014 | ||
1015 | r = dm_table_set_type(t); | |
1016 | if (r) { | |
1017 | DMERR("unable to set table type"); | |
1018 | return r; | |
1019 | } | |
1020 | ||
1021 | r = dm_table_build_index(t); | |
1022 | if (r) { | |
1023 | DMERR("unable to build btrees"); | |
1024 | return r; | |
1025 | } | |
1026 | ||
1027 | r = dm_table_prealloc_integrity(t, t->md); | |
1028 | if (r) { | |
1029 | DMERR("could not register integrity profile."); | |
1030 | return r; | |
1031 | } | |
1032 | ||
1033 | r = dm_table_alloc_md_mempools(t); | |
1034 | if (r) | |
1035 | DMERR("unable to allocate mempools"); | |
1036 | ||
1037 | return r; | |
1038 | } | |
1039 | ||
48c9c27b | 1040 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1041 | void dm_table_event_callback(struct dm_table *t, |
1042 | void (*fn)(void *), void *context) | |
1043 | { | |
48c9c27b | 1044 | mutex_lock(&_event_lock); |
1da177e4 LT |
1045 | t->event_fn = fn; |
1046 | t->event_context = context; | |
48c9c27b | 1047 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1048 | } |
1049 | ||
1050 | void dm_table_event(struct dm_table *t) | |
1051 | { | |
1052 | /* | |
1053 | * You can no longer call dm_table_event() from interrupt | |
1054 | * context, use a bottom half instead. | |
1055 | */ | |
1056 | BUG_ON(in_interrupt()); | |
1057 | ||
48c9c27b | 1058 | mutex_lock(&_event_lock); |
1da177e4 LT |
1059 | if (t->event_fn) |
1060 | t->event_fn(t->event_context); | |
48c9c27b | 1061 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1062 | } |
1063 | ||
1064 | sector_t dm_table_get_size(struct dm_table *t) | |
1065 | { | |
1066 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1067 | } | |
1068 | ||
1069 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1070 | { | |
14353539 | 1071 | if (index >= t->num_targets) |
1da177e4 LT |
1072 | return NULL; |
1073 | ||
1074 | return t->targets + index; | |
1075 | } | |
1076 | ||
1077 | /* | |
1078 | * Search the btree for the correct target. | |
512875bd JN |
1079 | * |
1080 | * Caller should check returned pointer with dm_target_is_valid() | |
1081 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1082 | */ |
1083 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1084 | { | |
1085 | unsigned int l, n = 0, k = 0; | |
1086 | sector_t *node; | |
1087 | ||
1088 | for (l = 0; l < t->depth; l++) { | |
1089 | n = get_child(n, k); | |
1090 | node = get_node(t, l, n); | |
1091 | ||
1092 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1093 | if (node[k] >= sector) | |
1094 | break; | |
1095 | } | |
1096 | ||
1097 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1098 | } | |
1099 | ||
754c5fc7 MS |
1100 | /* |
1101 | * Establish the new table's queue_limits and validate them. | |
1102 | */ | |
1103 | int dm_calculate_queue_limits(struct dm_table *table, | |
1104 | struct queue_limits *limits) | |
1105 | { | |
1106 | struct dm_target *uninitialized_var(ti); | |
1107 | struct queue_limits ti_limits; | |
1108 | unsigned i = 0; | |
1109 | ||
1110 | blk_set_default_limits(limits); | |
1111 | ||
1112 | while (i < dm_table_get_num_targets(table)) { | |
1113 | blk_set_default_limits(&ti_limits); | |
1114 | ||
1115 | ti = dm_table_get_target(table, i++); | |
1116 | ||
1117 | if (!ti->type->iterate_devices) | |
1118 | goto combine_limits; | |
1119 | ||
1120 | /* | |
1121 | * Combine queue limits of all the devices this target uses. | |
1122 | */ | |
1123 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1124 | &ti_limits); | |
1125 | ||
40bea431 MS |
1126 | /* Set I/O hints portion of queue limits */ |
1127 | if (ti->type->io_hints) | |
1128 | ti->type->io_hints(ti, &ti_limits); | |
1129 | ||
754c5fc7 MS |
1130 | /* |
1131 | * Check each device area is consistent with the target's | |
1132 | * overall queue limits. | |
1133 | */ | |
f6a1ed10 MP |
1134 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1135 | &ti_limits)) | |
754c5fc7 MS |
1136 | return -EINVAL; |
1137 | ||
1138 | combine_limits: | |
1139 | /* | |
1140 | * Merge this target's queue limits into the overall limits | |
1141 | * for the table. | |
1142 | */ | |
1143 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1144 | DMWARN("%s: adding target device " |
754c5fc7 | 1145 | "(start sect %llu len %llu) " |
b27d7f16 | 1146 | "caused an alignment inconsistency", |
754c5fc7 MS |
1147 | dm_device_name(table->md), |
1148 | (unsigned long long) ti->begin, | |
1149 | (unsigned long long) ti->len); | |
1150 | } | |
1151 | ||
1152 | return validate_hardware_logical_block_alignment(table, limits); | |
1153 | } | |
1154 | ||
9c47008d MP |
1155 | /* |
1156 | * Set the integrity profile for this device if all devices used have | |
a63a5cf8 MS |
1157 | * matching profiles. We're quite deep in the resume path but still |
1158 | * don't know if all devices (particularly DM devices this device | |
1159 | * may be stacked on) have matching profiles. Even if the profiles | |
1160 | * don't match we have no way to fail (to resume) at this point. | |
9c47008d MP |
1161 | */ |
1162 | static void dm_table_set_integrity(struct dm_table *t) | |
1163 | { | |
a63a5cf8 | 1164 | struct gendisk *template_disk = NULL; |
9c47008d MP |
1165 | |
1166 | if (!blk_get_integrity(dm_disk(t->md))) | |
1167 | return; | |
1168 | ||
a63a5cf8 MS |
1169 | template_disk = dm_table_get_integrity_disk(t, true); |
1170 | if (!template_disk && | |
1171 | blk_integrity_is_initialized(dm_disk(t->md))) { | |
1172 | DMWARN("%s: device no longer has a valid integrity profile", | |
1173 | dm_device_name(t->md)); | |
1174 | return; | |
9c47008d | 1175 | } |
9c47008d | 1176 | blk_integrity_register(dm_disk(t->md), |
a63a5cf8 | 1177 | blk_get_integrity(template_disk)); |
9c47008d MP |
1178 | } |
1179 | ||
754c5fc7 MS |
1180 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1181 | struct queue_limits *limits) | |
1da177e4 LT |
1182 | { |
1183 | /* | |
1197764e | 1184 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1185 | */ |
754c5fc7 | 1186 | q->limits = *limits; |
c9a3f6d6 | 1187 | |
5ae89a87 MS |
1188 | if (!dm_table_supports_discards(t)) |
1189 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1190 | else | |
1191 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1192 | ||
9c47008d | 1193 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1194 | |
1195 | /* | |
1196 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1197 | * visible to other CPUs because, once the flag is set, incoming bios | |
1198 | * are processed by request-based dm, which refers to the queue | |
1199 | * settings. | |
1200 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1201 | * md->deferred where queue settings are not needed yet. | |
1202 | * Those bios are passed to request-based dm at the resume time. | |
1203 | */ | |
1204 | smp_mb(); | |
1205 | if (dm_table_request_based(t)) | |
1206 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1207 | } |
1208 | ||
1209 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1210 | { | |
1211 | return t->num_targets; | |
1212 | } | |
1213 | ||
1214 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1215 | { | |
1216 | return &t->devices; | |
1217 | } | |
1218 | ||
aeb5d727 | 1219 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1220 | { |
1221 | return t->mode; | |
1222 | } | |
1223 | ||
1224 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1225 | { | |
1226 | int i = t->num_targets; | |
1227 | struct dm_target *ti = t->targets; | |
1228 | ||
1229 | while (i--) { | |
1230 | if (postsuspend) { | |
1231 | if (ti->type->postsuspend) | |
1232 | ti->type->postsuspend(ti); | |
1233 | } else if (ti->type->presuspend) | |
1234 | ti->type->presuspend(ti); | |
1235 | ||
1236 | ti++; | |
1237 | } | |
1238 | } | |
1239 | ||
1240 | void dm_table_presuspend_targets(struct dm_table *t) | |
1241 | { | |
cf222b37 AK |
1242 | if (!t) |
1243 | return; | |
1244 | ||
e8488d08 | 1245 | suspend_targets(t, 0); |
1da177e4 LT |
1246 | } |
1247 | ||
1248 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1249 | { | |
cf222b37 AK |
1250 | if (!t) |
1251 | return; | |
1252 | ||
e8488d08 | 1253 | suspend_targets(t, 1); |
1da177e4 LT |
1254 | } |
1255 | ||
8757b776 | 1256 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1257 | { |
8757b776 MB |
1258 | int i, r = 0; |
1259 | ||
1260 | for (i = 0; i < t->num_targets; i++) { | |
1261 | struct dm_target *ti = t->targets + i; | |
1262 | ||
1263 | if (!ti->type->preresume) | |
1264 | continue; | |
1265 | ||
1266 | r = ti->type->preresume(ti); | |
1267 | if (r) | |
1268 | return r; | |
1269 | } | |
1da177e4 LT |
1270 | |
1271 | for (i = 0; i < t->num_targets; i++) { | |
1272 | struct dm_target *ti = t->targets + i; | |
1273 | ||
1274 | if (ti->type->resume) | |
1275 | ti->type->resume(ti); | |
1276 | } | |
8757b776 MB |
1277 | |
1278 | return 0; | |
1da177e4 LT |
1279 | } |
1280 | ||
9d357b07 N |
1281 | void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb) |
1282 | { | |
1283 | list_add(&cb->list, &t->target_callbacks); | |
1284 | } | |
1285 | EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks); | |
1286 | ||
1da177e4 LT |
1287 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) |
1288 | { | |
82b1519b | 1289 | struct dm_dev_internal *dd; |
afb24528 | 1290 | struct list_head *devices = dm_table_get_devices(t); |
9d357b07 | 1291 | struct dm_target_callbacks *cb; |
1da177e4 LT |
1292 | int r = 0; |
1293 | ||
afb24528 | 1294 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1295 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1296 | char b[BDEVNAME_SIZE]; |
1297 | ||
1298 | if (likely(q)) | |
1299 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1300 | else | |
1301 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1302 | dm_device_name(t->md), | |
1303 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1304 | } |
1305 | ||
9d357b07 N |
1306 | list_for_each_entry(cb, &t->target_callbacks, list) |
1307 | if (cb->congested_fn) | |
1308 | r |= cb->congested_fn(cb, bdi_bits); | |
1309 | ||
1da177e4 LT |
1310 | return r; |
1311 | } | |
1312 | ||
cec47e3d KU |
1313 | int dm_table_any_busy_target(struct dm_table *t) |
1314 | { | |
1315 | unsigned i; | |
1316 | struct dm_target *ti; | |
1317 | ||
1318 | for (i = 0; i < t->num_targets; i++) { | |
1319 | ti = t->targets + i; | |
1320 | if (ti->type->busy && ti->type->busy(ti)) | |
1321 | return 1; | |
1322 | } | |
1323 | ||
1324 | return 0; | |
1325 | } | |
1326 | ||
1134e5ae MA |
1327 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1328 | { | |
1134e5ae MA |
1329 | return t->md; |
1330 | } | |
1331 | ||
5ae89a87 MS |
1332 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1333 | sector_t start, sector_t len, void *data) | |
1334 | { | |
1335 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1336 | ||
1337 | return q && blk_queue_discard(q); | |
1338 | } | |
1339 | ||
1340 | bool dm_table_supports_discards(struct dm_table *t) | |
1341 | { | |
1342 | struct dm_target *ti; | |
1343 | unsigned i = 0; | |
1344 | ||
1345 | if (!t->discards_supported) | |
1346 | return 0; | |
1347 | ||
1348 | /* | |
4c259327 MS |
1349 | * Unless any target used by the table set discards_supported, |
1350 | * require at least one underlying device to support discards. | |
5ae89a87 MS |
1351 | * t->devices includes internal dm devices such as mirror logs |
1352 | * so we need to use iterate_devices here, which targets | |
1353 | * supporting discard must provide. | |
1354 | */ | |
1355 | while (i < dm_table_get_num_targets(t)) { | |
1356 | ti = dm_table_get_target(t, i++); | |
1357 | ||
4c259327 MS |
1358 | if (ti->discards_supported) |
1359 | return 1; | |
1360 | ||
5ae89a87 MS |
1361 | if (ti->type->iterate_devices && |
1362 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1363 | return 1; | |
1364 | } | |
1365 | ||
1366 | return 0; | |
1367 | } | |
1368 | ||
1da177e4 LT |
1369 | EXPORT_SYMBOL(dm_vcalloc); |
1370 | EXPORT_SYMBOL(dm_get_device); | |
1371 | EXPORT_SYMBOL(dm_put_device); | |
1372 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1373 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1374 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1375 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1376 | EXPORT_SYMBOL(dm_table_put); |
1377 | EXPORT_SYMBOL(dm_table_get); |