]>
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 | { |
f4808ca9 | 365 | struct request_queue *q; |
754c5fc7 MS |
366 | struct queue_limits *limits = data; |
367 | struct block_device *bdev = dev->bdev; | |
368 | sector_t dev_size = | |
369 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 370 | unsigned short logical_block_size_sectors = |
754c5fc7 | 371 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 372 | char b[BDEVNAME_SIZE]; |
2cd54d9b | 373 | |
f4808ca9 MB |
374 | /* |
375 | * Some devices exist without request functions, | |
376 | * such as loop devices not yet bound to backing files. | |
377 | * Forbid the use of such devices. | |
378 | */ | |
379 | q = bdev_get_queue(bdev); | |
380 | if (!q || !q->make_request_fn) { | |
381 | DMWARN("%s: %s is not yet initialised: " | |
382 | "start=%llu, len=%llu, dev_size=%llu", | |
383 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
384 | (unsigned long long)start, | |
385 | (unsigned long long)len, | |
386 | (unsigned long long)dev_size); | |
387 | return 1; | |
388 | } | |
389 | ||
2cd54d9b | 390 | if (!dev_size) |
f6a1ed10 | 391 | return 0; |
2cd54d9b | 392 | |
5dea271b | 393 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
394 | DMWARN("%s: %s too small for target: " |
395 | "start=%llu, len=%llu, dev_size=%llu", | |
396 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
397 | (unsigned long long)start, | |
398 | (unsigned long long)len, | |
399 | (unsigned long long)dev_size); | |
f6a1ed10 | 400 | return 1; |
02acc3a4 MS |
401 | } |
402 | ||
403 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 404 | return 0; |
02acc3a4 MS |
405 | |
406 | if (start & (logical_block_size_sectors - 1)) { | |
407 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 408 | "logical block size %u of %s", |
02acc3a4 MS |
409 | dm_device_name(ti->table->md), |
410 | (unsigned long long)start, | |
754c5fc7 | 411 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 412 | return 1; |
02acc3a4 MS |
413 | } |
414 | ||
5dea271b | 415 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 416 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 417 | "logical block size %u of %s", |
02acc3a4 | 418 | dm_device_name(ti->table->md), |
5dea271b | 419 | (unsigned long long)len, |
754c5fc7 | 420 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 421 | return 1; |
02acc3a4 MS |
422 | } |
423 | ||
f6a1ed10 | 424 | return 0; |
1da177e4 LT |
425 | } |
426 | ||
427 | /* | |
570b9d96 | 428 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 429 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
430 | * device and not to touch the existing bdev field in case |
431 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 432 | */ |
aeb5d727 | 433 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 434 | struct mapped_device *md) |
1da177e4 LT |
435 | { |
436 | int r; | |
570b9d96 | 437 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 438 | |
570b9d96 AK |
439 | dd_new = dd_old = *dd; |
440 | ||
441 | dd_new.dm_dev.mode |= new_mode; | |
442 | dd_new.dm_dev.bdev = NULL; | |
443 | ||
444 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
445 | if (r) | |
446 | return r; | |
1da177e4 | 447 | |
82b1519b | 448 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 449 | close_dev(&dd_old, md); |
1da177e4 | 450 | |
570b9d96 | 451 | return 0; |
1da177e4 LT |
452 | } |
453 | ||
454 | /* | |
455 | * Add a device to the list, or just increment the usage count if | |
456 | * it's already present. | |
457 | */ | |
458 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
8215d6ec | 459 | const char *path, fmode_t mode, struct dm_dev **result) |
1da177e4 LT |
460 | { |
461 | int r; | |
69a2ce72 | 462 | dev_t uninitialized_var(dev); |
82b1519b | 463 | struct dm_dev_internal *dd; |
1da177e4 LT |
464 | unsigned int major, minor; |
465 | ||
547bc926 | 466 | BUG_ON(!t); |
1da177e4 LT |
467 | |
468 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
469 | /* Extract the major/minor numbers */ | |
470 | dev = MKDEV(major, minor); | |
471 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
472 | return -EOVERFLOW; | |
473 | } else { | |
474 | /* convert the path to a device */ | |
72e8264e CH |
475 | struct block_device *bdev = lookup_bdev(path); |
476 | ||
477 | if (IS_ERR(bdev)) | |
478 | return PTR_ERR(bdev); | |
479 | dev = bdev->bd_dev; | |
480 | bdput(bdev); | |
1da177e4 LT |
481 | } |
482 | ||
483 | dd = find_device(&t->devices, dev); | |
484 | if (!dd) { | |
485 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
486 | if (!dd) | |
487 | return -ENOMEM; | |
488 | ||
82b1519b MP |
489 | dd->dm_dev.mode = mode; |
490 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 491 | |
f165921d | 492 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
493 | kfree(dd); |
494 | return r; | |
495 | } | |
496 | ||
82b1519b | 497 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
498 | |
499 | atomic_set(&dd->count, 0); | |
500 | list_add(&dd->list, &t->devices); | |
501 | ||
82b1519b | 502 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 503 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
504 | if (r) |
505 | return r; | |
506 | } | |
507 | atomic_inc(&dd->count); | |
508 | ||
82b1519b | 509 | *result = &dd->dm_dev; |
1da177e4 LT |
510 | return 0; |
511 | } | |
512 | ||
754c5fc7 | 513 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 514 | sector_t start, sector_t len, void *data) |
1da177e4 | 515 | { |
754c5fc7 MS |
516 | struct queue_limits *limits = data; |
517 | struct block_device *bdev = dev->bdev; | |
165125e1 | 518 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
519 | char b[BDEVNAME_SIZE]; |
520 | ||
521 | if (unlikely(!q)) { | |
522 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
523 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 524 | return 0; |
0c2322e4 | 525 | } |
3cb40214 | 526 | |
b27d7f16 MP |
527 | if (bdev_stack_limits(limits, bdev, start) < 0) |
528 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
529 | "physical_block_size=%u, logical_block_size=%u, " |
530 | "alignment_offset=%u, start=%llu", | |
531 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
532 | q->limits.physical_block_size, | |
533 | q->limits.logical_block_size, | |
534 | q->limits.alignment_offset, | |
b27d7f16 | 535 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 536 | |
9980c638 MB |
537 | /* |
538 | * Check if merge fn is supported. | |
539 | * If not we'll force DM to use PAGE_SIZE or | |
540 | * smaller I/O, just to be safe. | |
3cb40214 | 541 | */ |
9980c638 MB |
542 | |
543 | if (q->merge_bvec_fn && !ti->type->merge) | |
72d4cd9f MS |
544 | blk_limits_max_hw_sectors(limits, |
545 | (unsigned int) (PAGE_SIZE >> 9)); | |
754c5fc7 | 546 | return 0; |
3cb40214 BR |
547 | } |
548 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 549 | |
8215d6ec NK |
550 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
551 | struct dm_dev **result) | |
3cb40214 | 552 | { |
8215d6ec | 553 | return __table_get_device(ti->table, ti, path, mode, result); |
1da177e4 LT |
554 | } |
555 | ||
754c5fc7 | 556 | |
1da177e4 LT |
557 | /* |
558 | * Decrement a devices use count and remove it if necessary. | |
559 | */ | |
82b1519b | 560 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 561 | { |
82b1519b MP |
562 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
563 | dm_dev); | |
564 | ||
1da177e4 | 565 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 566 | close_dev(dd, ti->table->md); |
1da177e4 LT |
567 | list_del(&dd->list); |
568 | kfree(dd); | |
569 | } | |
570 | } | |
571 | ||
572 | /* | |
573 | * Checks to see if the target joins onto the end of the table. | |
574 | */ | |
575 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
576 | { | |
577 | struct dm_target *prev; | |
578 | ||
579 | if (!table->num_targets) | |
580 | return !ti->begin; | |
581 | ||
582 | prev = &table->targets[table->num_targets - 1]; | |
583 | return (ti->begin == (prev->begin + prev->len)); | |
584 | } | |
585 | ||
586 | /* | |
587 | * Used to dynamically allocate the arg array. | |
588 | */ | |
589 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
590 | { | |
591 | char **argv; | |
592 | unsigned new_size; | |
593 | ||
594 | new_size = *array_size ? *array_size * 2 : 64; | |
595 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
596 | if (argv) { | |
597 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
598 | *array_size = new_size; | |
599 | } | |
600 | ||
601 | kfree(old_argv); | |
602 | return argv; | |
603 | } | |
604 | ||
605 | /* | |
606 | * Destructively splits up the argument list to pass to ctr. | |
607 | */ | |
608 | int dm_split_args(int *argc, char ***argvp, char *input) | |
609 | { | |
610 | char *start, *end = input, *out, **argv = NULL; | |
611 | unsigned array_size = 0; | |
612 | ||
613 | *argc = 0; | |
814d6862 DT |
614 | |
615 | if (!input) { | |
616 | *argvp = NULL; | |
617 | return 0; | |
618 | } | |
619 | ||
1da177e4 LT |
620 | argv = realloc_argv(&array_size, argv); |
621 | if (!argv) | |
622 | return -ENOMEM; | |
623 | ||
624 | while (1) { | |
1da177e4 | 625 | /* Skip whitespace */ |
e7d2860b | 626 | start = skip_spaces(end); |
1da177e4 LT |
627 | |
628 | if (!*start) | |
629 | break; /* success, we hit the end */ | |
630 | ||
631 | /* 'out' is used to remove any back-quotes */ | |
632 | end = out = start; | |
633 | while (*end) { | |
634 | /* Everything apart from '\0' can be quoted */ | |
635 | if (*end == '\\' && *(end + 1)) { | |
636 | *out++ = *(end + 1); | |
637 | end += 2; | |
638 | continue; | |
639 | } | |
640 | ||
641 | if (isspace(*end)) | |
642 | break; /* end of token */ | |
643 | ||
644 | *out++ = *end++; | |
645 | } | |
646 | ||
647 | /* have we already filled the array ? */ | |
648 | if ((*argc + 1) > array_size) { | |
649 | argv = realloc_argv(&array_size, argv); | |
650 | if (!argv) | |
651 | return -ENOMEM; | |
652 | } | |
653 | ||
654 | /* we know this is whitespace */ | |
655 | if (*end) | |
656 | end++; | |
657 | ||
658 | /* terminate the string and put it in the array */ | |
659 | *out = '\0'; | |
660 | argv[*argc] = start; | |
661 | (*argc)++; | |
662 | } | |
663 | ||
664 | *argvp = argv; | |
665 | return 0; | |
666 | } | |
667 | ||
be6d4305 MS |
668 | /* |
669 | * Impose necessary and sufficient conditions on a devices's table such | |
670 | * that any incoming bio which respects its logical_block_size can be | |
671 | * processed successfully. If it falls across the boundary between | |
672 | * two or more targets, the size of each piece it gets split into must | |
673 | * be compatible with the logical_block_size of the target processing it. | |
674 | */ | |
754c5fc7 MS |
675 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
676 | struct queue_limits *limits) | |
be6d4305 MS |
677 | { |
678 | /* | |
679 | * This function uses arithmetic modulo the logical_block_size | |
680 | * (in units of 512-byte sectors). | |
681 | */ | |
682 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 683 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
684 | |
685 | /* | |
686 | * Offset of the start of the next table entry, mod logical_block_size. | |
687 | */ | |
688 | unsigned short next_target_start = 0; | |
689 | ||
690 | /* | |
691 | * Given an aligned bio that extends beyond the end of a | |
692 | * target, how many sectors must the next target handle? | |
693 | */ | |
694 | unsigned short remaining = 0; | |
695 | ||
696 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 697 | struct queue_limits ti_limits; |
be6d4305 MS |
698 | unsigned i = 0; |
699 | ||
700 | /* | |
701 | * Check each entry in the table in turn. | |
702 | */ | |
703 | while (i < dm_table_get_num_targets(table)) { | |
704 | ti = dm_table_get_target(table, i++); | |
705 | ||
754c5fc7 MS |
706 | blk_set_default_limits(&ti_limits); |
707 | ||
708 | /* combine all target devices' limits */ | |
709 | if (ti->type->iterate_devices) | |
710 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
711 | &ti_limits); | |
712 | ||
be6d4305 MS |
713 | /* |
714 | * If the remaining sectors fall entirely within this | |
715 | * table entry are they compatible with its logical_block_size? | |
716 | */ | |
717 | if (remaining < ti->len && | |
754c5fc7 | 718 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
719 | SECTOR_SHIFT) - 1)) |
720 | break; /* Error */ | |
721 | ||
722 | next_target_start = | |
723 | (unsigned short) ((next_target_start + ti->len) & | |
724 | (device_logical_block_size_sects - 1)); | |
725 | remaining = next_target_start ? | |
726 | device_logical_block_size_sects - next_target_start : 0; | |
727 | } | |
728 | ||
729 | if (remaining) { | |
730 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 731 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
732 | dm_device_name(table->md), i, |
733 | (unsigned long long) ti->begin, | |
734 | (unsigned long long) ti->len, | |
754c5fc7 | 735 | limits->logical_block_size); |
be6d4305 MS |
736 | return -EINVAL; |
737 | } | |
738 | ||
739 | return 0; | |
740 | } | |
741 | ||
1da177e4 LT |
742 | int dm_table_add_target(struct dm_table *t, const char *type, |
743 | sector_t start, sector_t len, char *params) | |
744 | { | |
745 | int r = -EINVAL, argc; | |
746 | char **argv; | |
747 | struct dm_target *tgt; | |
748 | ||
749 | if ((r = check_space(t))) | |
750 | return r; | |
751 | ||
752 | tgt = t->targets + t->num_targets; | |
753 | memset(tgt, 0, sizeof(*tgt)); | |
754 | ||
755 | if (!len) { | |
72d94861 | 756 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
757 | return -EINVAL; |
758 | } | |
759 | ||
760 | tgt->type = dm_get_target_type(type); | |
761 | if (!tgt->type) { | |
72d94861 AK |
762 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
763 | type); | |
1da177e4 LT |
764 | return -EINVAL; |
765 | } | |
766 | ||
767 | tgt->table = t; | |
768 | tgt->begin = start; | |
769 | tgt->len = len; | |
770 | tgt->error = "Unknown error"; | |
771 | ||
772 | /* | |
773 | * Does this target adjoin the previous one ? | |
774 | */ | |
775 | if (!adjoin(t, tgt)) { | |
776 | tgt->error = "Gap in table"; | |
777 | r = -EINVAL; | |
778 | goto bad; | |
779 | } | |
780 | ||
781 | r = dm_split_args(&argc, &argv, params); | |
782 | if (r) { | |
783 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
784 | goto bad; | |
785 | } | |
786 | ||
787 | r = tgt->type->ctr(tgt, argc, argv); | |
788 | kfree(argv); | |
789 | if (r) | |
790 | goto bad; | |
791 | ||
792 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
793 | ||
5ae89a87 MS |
794 | if (!tgt->num_discard_requests) |
795 | t->discards_supported = 0; | |
796 | ||
1da177e4 LT |
797 | return 0; |
798 | ||
799 | bad: | |
72d94861 | 800 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
801 | dm_put_target_type(tgt->type); |
802 | return r; | |
803 | } | |
804 | ||
26803b9f | 805 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
806 | { |
807 | unsigned i; | |
808 | unsigned bio_based = 0, request_based = 0; | |
809 | struct dm_target *tgt; | |
810 | struct dm_dev_internal *dd; | |
811 | struct list_head *devices; | |
812 | ||
813 | for (i = 0; i < t->num_targets; i++) { | |
814 | tgt = t->targets + i; | |
815 | if (dm_target_request_based(tgt)) | |
816 | request_based = 1; | |
817 | else | |
818 | bio_based = 1; | |
819 | ||
820 | if (bio_based && request_based) { | |
821 | DMWARN("Inconsistent table: different target types" | |
822 | " can't be mixed up"); | |
823 | return -EINVAL; | |
824 | } | |
825 | } | |
826 | ||
827 | if (bio_based) { | |
828 | /* We must use this table as bio-based */ | |
829 | t->type = DM_TYPE_BIO_BASED; | |
830 | return 0; | |
831 | } | |
832 | ||
833 | BUG_ON(!request_based); /* No targets in this table */ | |
834 | ||
835 | /* Non-request-stackable devices can't be used for request-based dm */ | |
836 | devices = dm_table_get_devices(t); | |
837 | list_for_each_entry(dd, devices, list) { | |
838 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
839 | DMWARN("table load rejected: including" | |
840 | " non-request-stackable devices"); | |
841 | return -EINVAL; | |
842 | } | |
843 | } | |
844 | ||
845 | /* | |
846 | * Request-based dm supports only tables that have a single target now. | |
847 | * To support multiple targets, request splitting support is needed, | |
848 | * and that needs lots of changes in the block-layer. | |
849 | * (e.g. request completion process for partial completion.) | |
850 | */ | |
851 | if (t->num_targets > 1) { | |
852 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
853 | return -EINVAL; | |
854 | } | |
855 | ||
856 | t->type = DM_TYPE_REQUEST_BASED; | |
857 | ||
858 | return 0; | |
859 | } | |
860 | ||
861 | unsigned dm_table_get_type(struct dm_table *t) | |
862 | { | |
863 | return t->type; | |
864 | } | |
865 | ||
866 | bool dm_table_request_based(struct dm_table *t) | |
867 | { | |
868 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
869 | } | |
870 | ||
871 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
872 | { | |
873 | unsigned type = dm_table_get_type(t); | |
874 | ||
875 | if (unlikely(type == DM_TYPE_NONE)) { | |
876 | DMWARN("no table type is set, can't allocate mempools"); | |
877 | return -EINVAL; | |
878 | } | |
879 | ||
a91a2785 | 880 | t->mempools = dm_alloc_md_mempools(type, t->integrity_supported); |
e6ee8c0b KU |
881 | if (!t->mempools) |
882 | return -ENOMEM; | |
883 | ||
884 | return 0; | |
885 | } | |
886 | ||
887 | void dm_table_free_md_mempools(struct dm_table *t) | |
888 | { | |
889 | dm_free_md_mempools(t->mempools); | |
890 | t->mempools = NULL; | |
891 | } | |
892 | ||
893 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
894 | { | |
895 | return t->mempools; | |
896 | } | |
897 | ||
1da177e4 LT |
898 | static int setup_indexes(struct dm_table *t) |
899 | { | |
900 | int i; | |
901 | unsigned int total = 0; | |
902 | sector_t *indexes; | |
903 | ||
904 | /* allocate the space for *all* the indexes */ | |
905 | for (i = t->depth - 2; i >= 0; i--) { | |
906 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
907 | total += t->counts[i]; | |
908 | } | |
909 | ||
910 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
911 | if (!indexes) | |
912 | return -ENOMEM; | |
913 | ||
914 | /* set up internal nodes, bottom-up */ | |
82d601dc | 915 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
916 | t->index[i] = indexes; |
917 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
918 | setup_btree_index(i, t); | |
919 | } | |
920 | ||
921 | return 0; | |
922 | } | |
923 | ||
924 | /* | |
925 | * Builds the btree to index the map. | |
926 | */ | |
26803b9f | 927 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
928 | { |
929 | int r = 0; | |
930 | unsigned int leaf_nodes; | |
931 | ||
1da177e4 LT |
932 | /* how many indexes will the btree have ? */ |
933 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
934 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
935 | ||
936 | /* leaf layer has already been set up */ | |
937 | t->counts[t->depth - 1] = leaf_nodes; | |
938 | t->index[t->depth - 1] = t->highs; | |
939 | ||
940 | if (t->depth >= 2) | |
941 | r = setup_indexes(t); | |
942 | ||
943 | return r; | |
944 | } | |
945 | ||
a63a5cf8 MS |
946 | /* |
947 | * Get a disk whose integrity profile reflects the table's profile. | |
948 | * If %match_all is true, all devices' profiles must match. | |
949 | * If %match_all is false, all devices must at least have an | |
950 | * allocated integrity profile; but uninitialized is ok. | |
951 | * Returns NULL if integrity support was inconsistent or unavailable. | |
952 | */ | |
953 | static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t, | |
954 | bool match_all) | |
955 | { | |
956 | struct list_head *devices = dm_table_get_devices(t); | |
957 | struct dm_dev_internal *dd = NULL; | |
958 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
959 | ||
960 | list_for_each_entry(dd, devices, list) { | |
961 | template_disk = dd->dm_dev.bdev->bd_disk; | |
962 | if (!blk_get_integrity(template_disk)) | |
963 | goto no_integrity; | |
964 | if (!match_all && !blk_integrity_is_initialized(template_disk)) | |
965 | continue; /* skip uninitialized profiles */ | |
966 | else if (prev_disk && | |
967 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
968 | goto no_integrity; | |
969 | prev_disk = template_disk; | |
970 | } | |
971 | ||
972 | return template_disk; | |
973 | ||
974 | no_integrity: | |
975 | if (prev_disk) | |
976 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
977 | dm_device_name(t->md), | |
978 | prev_disk->disk_name, | |
979 | template_disk->disk_name); | |
980 | return NULL; | |
981 | } | |
982 | ||
26803b9f WD |
983 | /* |
984 | * Register the mapped device for blk_integrity support if | |
a63a5cf8 MS |
985 | * the underlying devices have an integrity profile. But all devices |
986 | * may not have matching profiles (checking all devices isn't reliable | |
987 | * during table load because this table may use other DM device(s) which | |
988 | * must be resumed before they will have an initialized integity profile). | |
989 | * Stacked DM devices force a 2 stage integrity profile validation: | |
990 | * 1 - during load, validate all initialized integrity profiles match | |
991 | * 2 - during resume, validate all integrity profiles match | |
26803b9f WD |
992 | */ |
993 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
994 | { | |
a63a5cf8 | 995 | struct gendisk *template_disk = NULL; |
26803b9f | 996 | |
a63a5cf8 MS |
997 | template_disk = dm_table_get_integrity_disk(t, false); |
998 | if (!template_disk) | |
999 | return 0; | |
26803b9f | 1000 | |
a63a5cf8 MS |
1001 | if (!blk_integrity_is_initialized(dm_disk(md))) { |
1002 | t->integrity_supported = 1; | |
1003 | return blk_integrity_register(dm_disk(md), NULL); | |
1004 | } | |
1005 | ||
1006 | /* | |
1007 | * If DM device already has an initalized integrity | |
1008 | * profile the new profile should not conflict. | |
1009 | */ | |
1010 | if (blk_integrity_is_initialized(template_disk) && | |
1011 | blk_integrity_compare(dm_disk(md), template_disk) < 0) { | |
1012 | DMWARN("%s: conflict with existing integrity profile: " | |
1013 | "%s profile mismatch", | |
1014 | dm_device_name(t->md), | |
1015 | template_disk->disk_name); | |
1016 | return 1; | |
1017 | } | |
1018 | ||
1019 | /* Preserve existing initialized integrity profile */ | |
1020 | t->integrity_supported = 1; | |
26803b9f WD |
1021 | return 0; |
1022 | } | |
1023 | ||
1024 | /* | |
1025 | * Prepares the table for use by building the indices, | |
1026 | * setting the type, and allocating mempools. | |
1027 | */ | |
1028 | int dm_table_complete(struct dm_table *t) | |
1029 | { | |
1030 | int r; | |
1031 | ||
1032 | r = dm_table_set_type(t); | |
1033 | if (r) { | |
1034 | DMERR("unable to set table type"); | |
1035 | return r; | |
1036 | } | |
1037 | ||
1038 | r = dm_table_build_index(t); | |
1039 | if (r) { | |
1040 | DMERR("unable to build btrees"); | |
1041 | return r; | |
1042 | } | |
1043 | ||
1044 | r = dm_table_prealloc_integrity(t, t->md); | |
1045 | if (r) { | |
1046 | DMERR("could not register integrity profile."); | |
1047 | return r; | |
1048 | } | |
1049 | ||
1050 | r = dm_table_alloc_md_mempools(t); | |
1051 | if (r) | |
1052 | DMERR("unable to allocate mempools"); | |
1053 | ||
1054 | return r; | |
1055 | } | |
1056 | ||
48c9c27b | 1057 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1058 | void dm_table_event_callback(struct dm_table *t, |
1059 | void (*fn)(void *), void *context) | |
1060 | { | |
48c9c27b | 1061 | mutex_lock(&_event_lock); |
1da177e4 LT |
1062 | t->event_fn = fn; |
1063 | t->event_context = context; | |
48c9c27b | 1064 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1065 | } |
1066 | ||
1067 | void dm_table_event(struct dm_table *t) | |
1068 | { | |
1069 | /* | |
1070 | * You can no longer call dm_table_event() from interrupt | |
1071 | * context, use a bottom half instead. | |
1072 | */ | |
1073 | BUG_ON(in_interrupt()); | |
1074 | ||
48c9c27b | 1075 | mutex_lock(&_event_lock); |
1da177e4 LT |
1076 | if (t->event_fn) |
1077 | t->event_fn(t->event_context); | |
48c9c27b | 1078 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1079 | } |
1080 | ||
1081 | sector_t dm_table_get_size(struct dm_table *t) | |
1082 | { | |
1083 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1084 | } | |
1085 | ||
1086 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1087 | { | |
14353539 | 1088 | if (index >= t->num_targets) |
1da177e4 LT |
1089 | return NULL; |
1090 | ||
1091 | return t->targets + index; | |
1092 | } | |
1093 | ||
1094 | /* | |
1095 | * Search the btree for the correct target. | |
512875bd JN |
1096 | * |
1097 | * Caller should check returned pointer with dm_target_is_valid() | |
1098 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1099 | */ |
1100 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1101 | { | |
1102 | unsigned int l, n = 0, k = 0; | |
1103 | sector_t *node; | |
1104 | ||
1105 | for (l = 0; l < t->depth; l++) { | |
1106 | n = get_child(n, k); | |
1107 | node = get_node(t, l, n); | |
1108 | ||
1109 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1110 | if (node[k] >= sector) | |
1111 | break; | |
1112 | } | |
1113 | ||
1114 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1115 | } | |
1116 | ||
754c5fc7 MS |
1117 | /* |
1118 | * Establish the new table's queue_limits and validate them. | |
1119 | */ | |
1120 | int dm_calculate_queue_limits(struct dm_table *table, | |
1121 | struct queue_limits *limits) | |
1122 | { | |
1123 | struct dm_target *uninitialized_var(ti); | |
1124 | struct queue_limits ti_limits; | |
1125 | unsigned i = 0; | |
1126 | ||
1127 | blk_set_default_limits(limits); | |
1128 | ||
1129 | while (i < dm_table_get_num_targets(table)) { | |
1130 | blk_set_default_limits(&ti_limits); | |
1131 | ||
1132 | ti = dm_table_get_target(table, i++); | |
1133 | ||
1134 | if (!ti->type->iterate_devices) | |
1135 | goto combine_limits; | |
1136 | ||
1137 | /* | |
1138 | * Combine queue limits of all the devices this target uses. | |
1139 | */ | |
1140 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1141 | &ti_limits); | |
1142 | ||
40bea431 MS |
1143 | /* Set I/O hints portion of queue limits */ |
1144 | if (ti->type->io_hints) | |
1145 | ti->type->io_hints(ti, &ti_limits); | |
1146 | ||
754c5fc7 MS |
1147 | /* |
1148 | * Check each device area is consistent with the target's | |
1149 | * overall queue limits. | |
1150 | */ | |
f6a1ed10 MP |
1151 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1152 | &ti_limits)) | |
754c5fc7 MS |
1153 | return -EINVAL; |
1154 | ||
1155 | combine_limits: | |
1156 | /* | |
1157 | * Merge this target's queue limits into the overall limits | |
1158 | * for the table. | |
1159 | */ | |
1160 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1161 | DMWARN("%s: adding target device " |
754c5fc7 | 1162 | "(start sect %llu len %llu) " |
b27d7f16 | 1163 | "caused an alignment inconsistency", |
754c5fc7 MS |
1164 | dm_device_name(table->md), |
1165 | (unsigned long long) ti->begin, | |
1166 | (unsigned long long) ti->len); | |
1167 | } | |
1168 | ||
1169 | return validate_hardware_logical_block_alignment(table, limits); | |
1170 | } | |
1171 | ||
9c47008d MP |
1172 | /* |
1173 | * Set the integrity profile for this device if all devices used have | |
a63a5cf8 MS |
1174 | * matching profiles. We're quite deep in the resume path but still |
1175 | * don't know if all devices (particularly DM devices this device | |
1176 | * may be stacked on) have matching profiles. Even if the profiles | |
1177 | * don't match we have no way to fail (to resume) at this point. | |
9c47008d MP |
1178 | */ |
1179 | static void dm_table_set_integrity(struct dm_table *t) | |
1180 | { | |
a63a5cf8 | 1181 | struct gendisk *template_disk = NULL; |
9c47008d MP |
1182 | |
1183 | if (!blk_get_integrity(dm_disk(t->md))) | |
1184 | return; | |
1185 | ||
a63a5cf8 MS |
1186 | template_disk = dm_table_get_integrity_disk(t, true); |
1187 | if (!template_disk && | |
1188 | blk_integrity_is_initialized(dm_disk(t->md))) { | |
1189 | DMWARN("%s: device no longer has a valid integrity profile", | |
1190 | dm_device_name(t->md)); | |
1191 | return; | |
9c47008d | 1192 | } |
9c47008d | 1193 | blk_integrity_register(dm_disk(t->md), |
a63a5cf8 | 1194 | blk_get_integrity(template_disk)); |
9c47008d MP |
1195 | } |
1196 | ||
754c5fc7 MS |
1197 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1198 | struct queue_limits *limits) | |
1da177e4 LT |
1199 | { |
1200 | /* | |
1197764e | 1201 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1202 | */ |
754c5fc7 | 1203 | q->limits = *limits; |
c9a3f6d6 | 1204 | |
5ae89a87 MS |
1205 | if (!dm_table_supports_discards(t)) |
1206 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1207 | else | |
1208 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1209 | ||
9c47008d | 1210 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1211 | |
1212 | /* | |
1213 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1214 | * visible to other CPUs because, once the flag is set, incoming bios | |
1215 | * are processed by request-based dm, which refers to the queue | |
1216 | * settings. | |
1217 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1218 | * md->deferred where queue settings are not needed yet. | |
1219 | * Those bios are passed to request-based dm at the resume time. | |
1220 | */ | |
1221 | smp_mb(); | |
1222 | if (dm_table_request_based(t)) | |
1223 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1224 | } |
1225 | ||
1226 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1227 | { | |
1228 | return t->num_targets; | |
1229 | } | |
1230 | ||
1231 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1232 | { | |
1233 | return &t->devices; | |
1234 | } | |
1235 | ||
aeb5d727 | 1236 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1237 | { |
1238 | return t->mode; | |
1239 | } | |
1240 | ||
1241 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1242 | { | |
1243 | int i = t->num_targets; | |
1244 | struct dm_target *ti = t->targets; | |
1245 | ||
1246 | while (i--) { | |
1247 | if (postsuspend) { | |
1248 | if (ti->type->postsuspend) | |
1249 | ti->type->postsuspend(ti); | |
1250 | } else if (ti->type->presuspend) | |
1251 | ti->type->presuspend(ti); | |
1252 | ||
1253 | ti++; | |
1254 | } | |
1255 | } | |
1256 | ||
1257 | void dm_table_presuspend_targets(struct dm_table *t) | |
1258 | { | |
cf222b37 AK |
1259 | if (!t) |
1260 | return; | |
1261 | ||
e8488d08 | 1262 | suspend_targets(t, 0); |
1da177e4 LT |
1263 | } |
1264 | ||
1265 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1266 | { | |
cf222b37 AK |
1267 | if (!t) |
1268 | return; | |
1269 | ||
e8488d08 | 1270 | suspend_targets(t, 1); |
1da177e4 LT |
1271 | } |
1272 | ||
8757b776 | 1273 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1274 | { |
8757b776 MB |
1275 | int i, r = 0; |
1276 | ||
1277 | for (i = 0; i < t->num_targets; i++) { | |
1278 | struct dm_target *ti = t->targets + i; | |
1279 | ||
1280 | if (!ti->type->preresume) | |
1281 | continue; | |
1282 | ||
1283 | r = ti->type->preresume(ti); | |
1284 | if (r) | |
1285 | return r; | |
1286 | } | |
1da177e4 LT |
1287 | |
1288 | for (i = 0; i < t->num_targets; i++) { | |
1289 | struct dm_target *ti = t->targets + i; | |
1290 | ||
1291 | if (ti->type->resume) | |
1292 | ti->type->resume(ti); | |
1293 | } | |
8757b776 MB |
1294 | |
1295 | return 0; | |
1da177e4 LT |
1296 | } |
1297 | ||
9d357b07 N |
1298 | void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb) |
1299 | { | |
1300 | list_add(&cb->list, &t->target_callbacks); | |
1301 | } | |
1302 | EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks); | |
1303 | ||
1da177e4 LT |
1304 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) |
1305 | { | |
82b1519b | 1306 | struct dm_dev_internal *dd; |
afb24528 | 1307 | struct list_head *devices = dm_table_get_devices(t); |
9d357b07 | 1308 | struct dm_target_callbacks *cb; |
1da177e4 LT |
1309 | int r = 0; |
1310 | ||
afb24528 | 1311 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1312 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1313 | char b[BDEVNAME_SIZE]; |
1314 | ||
1315 | if (likely(q)) | |
1316 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1317 | else | |
1318 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1319 | dm_device_name(t->md), | |
1320 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1321 | } |
1322 | ||
9d357b07 N |
1323 | list_for_each_entry(cb, &t->target_callbacks, list) |
1324 | if (cb->congested_fn) | |
1325 | r |= cb->congested_fn(cb, bdi_bits); | |
1326 | ||
1da177e4 LT |
1327 | return r; |
1328 | } | |
1329 | ||
cec47e3d KU |
1330 | int dm_table_any_busy_target(struct dm_table *t) |
1331 | { | |
1332 | unsigned i; | |
1333 | struct dm_target *ti; | |
1334 | ||
1335 | for (i = 0; i < t->num_targets; i++) { | |
1336 | ti = t->targets + i; | |
1337 | if (ti->type->busy && ti->type->busy(ti)) | |
1338 | return 1; | |
1339 | } | |
1340 | ||
1341 | return 0; | |
1342 | } | |
1343 | ||
1134e5ae MA |
1344 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1345 | { | |
1134e5ae MA |
1346 | return t->md; |
1347 | } | |
1348 | ||
5ae89a87 MS |
1349 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1350 | sector_t start, sector_t len, void *data) | |
1351 | { | |
1352 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1353 | ||
1354 | return q && blk_queue_discard(q); | |
1355 | } | |
1356 | ||
1357 | bool dm_table_supports_discards(struct dm_table *t) | |
1358 | { | |
1359 | struct dm_target *ti; | |
1360 | unsigned i = 0; | |
1361 | ||
1362 | if (!t->discards_supported) | |
1363 | return 0; | |
1364 | ||
1365 | /* | |
4c259327 MS |
1366 | * Unless any target used by the table set discards_supported, |
1367 | * require at least one underlying device to support discards. | |
5ae89a87 MS |
1368 | * t->devices includes internal dm devices such as mirror logs |
1369 | * so we need to use iterate_devices here, which targets | |
1370 | * supporting discard must provide. | |
1371 | */ | |
1372 | while (i < dm_table_get_num_targets(t)) { | |
1373 | ti = dm_table_get_target(t, i++); | |
1374 | ||
4c259327 MS |
1375 | if (ti->discards_supported) |
1376 | return 1; | |
1377 | ||
5ae89a87 MS |
1378 | if (ti->type->iterate_devices && |
1379 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1380 | return 1; | |
1381 | } | |
1382 | ||
1383 | return 0; | |
1384 | } | |
1385 | ||
1da177e4 LT |
1386 | EXPORT_SYMBOL(dm_vcalloc); |
1387 | EXPORT_SYMBOL(dm_get_device); | |
1388 | EXPORT_SYMBOL(dm_put_device); | |
1389 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1390 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1391 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1392 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1393 | EXPORT_SYMBOL(dm_table_put); |
1394 | EXPORT_SYMBOL(dm_table_get); |