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> |
60063497 | 20 | #include <linux/atomic.h> |
1da177e4 | 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 | ||
36a0456f | 57 | struct target_type *immutable_target_type; |
a91a2785 | 58 | unsigned integrity_supported:1; |
3791e2fc | 59 | unsigned singleton:1; |
5ae89a87 | 60 | |
1da177e4 LT |
61 | /* |
62 | * Indicates the rw permissions for the new logical | |
63 | * device. This should be a combination of FMODE_READ | |
64 | * and FMODE_WRITE. | |
65 | */ | |
aeb5d727 | 66 | fmode_t mode; |
1da177e4 LT |
67 | |
68 | /* a list of devices used by this table */ | |
69 | struct list_head devices; | |
70 | ||
1da177e4 LT |
71 | /* events get handed up using this callback */ |
72 | void (*event_fn)(void *); | |
73 | void *event_context; | |
e6ee8c0b KU |
74 | |
75 | struct dm_md_mempools *mempools; | |
9d357b07 N |
76 | |
77 | struct list_head target_callbacks; | |
1da177e4 LT |
78 | }; |
79 | ||
80 | /* | |
81 | * Similar to ceiling(log_size(n)) | |
82 | */ | |
83 | static unsigned int int_log(unsigned int n, unsigned int base) | |
84 | { | |
85 | int result = 0; | |
86 | ||
87 | while (n > 1) { | |
88 | n = dm_div_up(n, base); | |
89 | result++; | |
90 | } | |
91 | ||
92 | return result; | |
93 | } | |
94 | ||
1da177e4 LT |
95 | /* |
96 | * Calculate the index of the child node of the n'th node k'th key. | |
97 | */ | |
98 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
99 | { | |
100 | return (n * CHILDREN_PER_NODE) + k; | |
101 | } | |
102 | ||
103 | /* | |
104 | * Return the n'th node of level l from table t. | |
105 | */ | |
106 | static inline sector_t *get_node(struct dm_table *t, | |
107 | unsigned int l, unsigned int n) | |
108 | { | |
109 | return t->index[l] + (n * KEYS_PER_NODE); | |
110 | } | |
111 | ||
112 | /* | |
113 | * Return the highest key that you could lookup from the n'th | |
114 | * node on level l of the btree. | |
115 | */ | |
116 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
117 | { | |
118 | for (; l < t->depth - 1; l++) | |
119 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
120 | ||
121 | if (n >= t->counts[l]) | |
122 | return (sector_t) - 1; | |
123 | ||
124 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
125 | } | |
126 | ||
127 | /* | |
128 | * Fills in a level of the btree based on the highs of the level | |
129 | * below it. | |
130 | */ | |
131 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
132 | { | |
133 | unsigned int n, k; | |
134 | sector_t *node; | |
135 | ||
136 | for (n = 0U; n < t->counts[l]; n++) { | |
137 | node = get_node(t, l, n); | |
138 | ||
139 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
140 | node[k] = high(t, l + 1, get_child(n, k)); | |
141 | } | |
142 | ||
143 | return 0; | |
144 | } | |
145 | ||
146 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
147 | { | |
148 | unsigned long size; | |
149 | void *addr; | |
150 | ||
151 | /* | |
152 | * Check that we're not going to overflow. | |
153 | */ | |
154 | if (nmemb > (ULONG_MAX / elem_size)) | |
155 | return NULL; | |
156 | ||
157 | size = nmemb * elem_size; | |
e29e65aa | 158 | addr = vzalloc(size); |
1da177e4 LT |
159 | |
160 | return addr; | |
161 | } | |
08649012 | 162 | EXPORT_SYMBOL(dm_vcalloc); |
1da177e4 LT |
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); |
1da177e4 LT |
212 | |
213 | if (!num_targets) | |
214 | num_targets = KEYS_PER_NODE; | |
215 | ||
216 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
217 | ||
218 | if (alloc_targets(t, num_targets)) { | |
219 | kfree(t); | |
220 | t = NULL; | |
221 | return -ENOMEM; | |
222 | } | |
223 | ||
224 | t->mode = mode; | |
1134e5ae | 225 | t->md = md; |
1da177e4 LT |
226 | *result = t; |
227 | return 0; | |
228 | } | |
229 | ||
230 | static void free_devices(struct list_head *devices) | |
231 | { | |
232 | struct list_head *tmp, *next; | |
233 | ||
afb24528 | 234 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
235 | struct dm_dev_internal *dd = |
236 | list_entry(tmp, struct dm_dev_internal, list); | |
1b6da754 JB |
237 | DMWARN("dm_table_destroy: dm_put_device call missing for %s", |
238 | dd->dm_dev.name); | |
1da177e4 LT |
239 | kfree(dd); |
240 | } | |
241 | } | |
242 | ||
d5816876 | 243 | void dm_table_destroy(struct dm_table *t) |
1da177e4 LT |
244 | { |
245 | unsigned int i; | |
246 | ||
a7940155 AK |
247 | if (!t) |
248 | return; | |
249 | ||
d5816876 MP |
250 | while (atomic_read(&t->holders)) |
251 | msleep(1); | |
252 | smp_mb(); | |
253 | ||
26803b9f | 254 | /* free the indexes */ |
1da177e4 LT |
255 | if (t->depth >= 2) |
256 | vfree(t->index[t->depth - 2]); | |
257 | ||
258 | /* free the targets */ | |
259 | for (i = 0; i < t->num_targets; i++) { | |
260 | struct dm_target *tgt = t->targets + i; | |
261 | ||
262 | if (tgt->type->dtr) | |
263 | tgt->type->dtr(tgt); | |
264 | ||
265 | dm_put_target_type(tgt->type); | |
266 | } | |
267 | ||
268 | vfree(t->highs); | |
269 | ||
270 | /* free the device list */ | |
1b6da754 | 271 | if (t->devices.next != &t->devices) |
1da177e4 | 272 | free_devices(&t->devices); |
1da177e4 | 273 | |
e6ee8c0b KU |
274 | dm_free_md_mempools(t->mempools); |
275 | ||
1da177e4 LT |
276 | kfree(t); |
277 | } | |
278 | ||
279 | void dm_table_get(struct dm_table *t) | |
280 | { | |
281 | atomic_inc(&t->holders); | |
282 | } | |
08649012 | 283 | EXPORT_SYMBOL(dm_table_get); |
1da177e4 LT |
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 | 292 | } |
08649012 | 293 | EXPORT_SYMBOL(dm_table_put); |
1da177e4 LT |
294 | |
295 | /* | |
296 | * Checks to see if we need to extend highs or targets. | |
297 | */ | |
298 | static inline int check_space(struct dm_table *t) | |
299 | { | |
300 | if (t->num_targets >= t->num_allocated) | |
301 | return alloc_targets(t, t->num_allocated * 2); | |
302 | ||
303 | return 0; | |
304 | } | |
305 | ||
1da177e4 LT |
306 | /* |
307 | * See if we've already got a device in the list. | |
308 | */ | |
82b1519b | 309 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 310 | { |
82b1519b | 311 | struct dm_dev_internal *dd; |
1da177e4 LT |
312 | |
313 | list_for_each_entry (dd, l, list) | |
82b1519b | 314 | if (dd->dm_dev.bdev->bd_dev == dev) |
1da177e4 LT |
315 | return dd; |
316 | ||
317 | return NULL; | |
318 | } | |
319 | ||
320 | /* | |
321 | * Open a device so we can use it as a map destination. | |
322 | */ | |
82b1519b MP |
323 | static int open_dev(struct dm_dev_internal *d, dev_t dev, |
324 | struct mapped_device *md) | |
1da177e4 LT |
325 | { |
326 | static char *_claim_ptr = "I belong to device-mapper"; | |
327 | struct block_device *bdev; | |
328 | ||
329 | int r; | |
330 | ||
82b1519b | 331 | BUG_ON(d->dm_dev.bdev); |
1da177e4 | 332 | |
d4d77629 | 333 | bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr); |
1da177e4 LT |
334 | if (IS_ERR(bdev)) |
335 | return PTR_ERR(bdev); | |
e09b457b | 336 | |
e09b457b TH |
337 | r = bd_link_disk_holder(bdev, dm_disk(md)); |
338 | if (r) { | |
e525fd89 | 339 | blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL); |
e09b457b TH |
340 | return r; |
341 | } | |
342 | ||
343 | d->dm_dev.bdev = bdev; | |
344 | return 0; | |
1da177e4 LT |
345 | } |
346 | ||
347 | /* | |
348 | * Close a device that we've been using. | |
349 | */ | |
82b1519b | 350 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 351 | { |
82b1519b | 352 | if (!d->dm_dev.bdev) |
1da177e4 LT |
353 | return; |
354 | ||
49731baa | 355 | bd_unlink_disk_holder(d->dm_dev.bdev, dm_disk(md)); |
e525fd89 | 356 | blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL); |
82b1519b | 357 | d->dm_dev.bdev = NULL; |
1da177e4 LT |
358 | } |
359 | ||
360 | /* | |
f6a1ed10 | 361 | * If possible, this checks an area of a destination device is invalid. |
1da177e4 | 362 | */ |
f6a1ed10 MP |
363 | static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, |
364 | sector_t start, sector_t len, void *data) | |
1da177e4 | 365 | { |
f4808ca9 | 366 | struct request_queue *q; |
754c5fc7 MS |
367 | struct queue_limits *limits = data; |
368 | struct block_device *bdev = dev->bdev; | |
369 | sector_t dev_size = | |
370 | i_size_read(bdev->bd_inode) >> SECTOR_SHIFT; | |
02acc3a4 | 371 | unsigned short logical_block_size_sectors = |
754c5fc7 | 372 | limits->logical_block_size >> SECTOR_SHIFT; |
02acc3a4 | 373 | char b[BDEVNAME_SIZE]; |
2cd54d9b | 374 | |
f4808ca9 MB |
375 | /* |
376 | * Some devices exist without request functions, | |
377 | * such as loop devices not yet bound to backing files. | |
378 | * Forbid the use of such devices. | |
379 | */ | |
380 | q = bdev_get_queue(bdev); | |
381 | if (!q || !q->make_request_fn) { | |
382 | DMWARN("%s: %s is not yet initialised: " | |
383 | "start=%llu, len=%llu, dev_size=%llu", | |
384 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
385 | (unsigned long long)start, | |
386 | (unsigned long long)len, | |
387 | (unsigned long long)dev_size); | |
388 | return 1; | |
389 | } | |
390 | ||
2cd54d9b | 391 | if (!dev_size) |
f6a1ed10 | 392 | return 0; |
2cd54d9b | 393 | |
5dea271b | 394 | if ((start >= dev_size) || (start + len > dev_size)) { |
a963a956 MS |
395 | DMWARN("%s: %s too small for target: " |
396 | "start=%llu, len=%llu, dev_size=%llu", | |
397 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
398 | (unsigned long long)start, | |
399 | (unsigned long long)len, | |
400 | (unsigned long long)dev_size); | |
f6a1ed10 | 401 | return 1; |
02acc3a4 MS |
402 | } |
403 | ||
404 | if (logical_block_size_sectors <= 1) | |
f6a1ed10 | 405 | return 0; |
02acc3a4 MS |
406 | |
407 | if (start & (logical_block_size_sectors - 1)) { | |
408 | DMWARN("%s: start=%llu not aligned to h/w " | |
a963a956 | 409 | "logical block size %u of %s", |
02acc3a4 MS |
410 | dm_device_name(ti->table->md), |
411 | (unsigned long long)start, | |
754c5fc7 | 412 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 413 | return 1; |
02acc3a4 MS |
414 | } |
415 | ||
5dea271b | 416 | if (len & (logical_block_size_sectors - 1)) { |
02acc3a4 | 417 | DMWARN("%s: len=%llu not aligned to h/w " |
a963a956 | 418 | "logical block size %u of %s", |
02acc3a4 | 419 | dm_device_name(ti->table->md), |
5dea271b | 420 | (unsigned long long)len, |
754c5fc7 | 421 | limits->logical_block_size, bdevname(bdev, b)); |
f6a1ed10 | 422 | return 1; |
02acc3a4 MS |
423 | } |
424 | ||
f6a1ed10 | 425 | return 0; |
1da177e4 LT |
426 | } |
427 | ||
428 | /* | |
570b9d96 | 429 | * This upgrades the mode on an already open dm_dev, being |
1da177e4 | 430 | * careful to leave things as they were if we fail to reopen the |
570b9d96 AK |
431 | * device and not to touch the existing bdev field in case |
432 | * it is accessed concurrently inside dm_table_any_congested(). | |
1da177e4 | 433 | */ |
aeb5d727 | 434 | static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, |
82b1519b | 435 | struct mapped_device *md) |
1da177e4 LT |
436 | { |
437 | int r; | |
570b9d96 | 438 | struct dm_dev_internal dd_new, dd_old; |
1da177e4 | 439 | |
570b9d96 AK |
440 | dd_new = dd_old = *dd; |
441 | ||
442 | dd_new.dm_dev.mode |= new_mode; | |
443 | dd_new.dm_dev.bdev = NULL; | |
444 | ||
445 | r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md); | |
446 | if (r) | |
447 | return r; | |
1da177e4 | 448 | |
82b1519b | 449 | dd->dm_dev.mode |= new_mode; |
570b9d96 | 450 | close_dev(&dd_old, md); |
1da177e4 | 451 | |
570b9d96 | 452 | return 0; |
1da177e4 LT |
453 | } |
454 | ||
455 | /* | |
456 | * Add a device to the list, or just increment the usage count if | |
457 | * it's already present. | |
458 | */ | |
08649012 MS |
459 | int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, |
460 | struct dm_dev **result) | |
1da177e4 LT |
461 | { |
462 | int r; | |
69a2ce72 | 463 | dev_t uninitialized_var(dev); |
82b1519b | 464 | struct dm_dev_internal *dd; |
1da177e4 | 465 | unsigned int major, minor; |
08649012 | 466 | struct dm_table *t = ti->table; |
1da177e4 | 467 | |
547bc926 | 468 | BUG_ON(!t); |
1da177e4 LT |
469 | |
470 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
471 | /* Extract the major/minor numbers */ | |
472 | dev = MKDEV(major, minor); | |
473 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
474 | return -EOVERFLOW; | |
475 | } else { | |
476 | /* convert the path to a device */ | |
72e8264e CH |
477 | struct block_device *bdev = lookup_bdev(path); |
478 | ||
479 | if (IS_ERR(bdev)) | |
480 | return PTR_ERR(bdev); | |
481 | dev = bdev->bd_dev; | |
482 | bdput(bdev); | |
1da177e4 LT |
483 | } |
484 | ||
485 | dd = find_device(&t->devices, dev); | |
486 | if (!dd) { | |
487 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
488 | if (!dd) | |
489 | return -ENOMEM; | |
490 | ||
82b1519b MP |
491 | dd->dm_dev.mode = mode; |
492 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 493 | |
f165921d | 494 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
495 | kfree(dd); |
496 | return r; | |
497 | } | |
498 | ||
82b1519b | 499 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
500 | |
501 | atomic_set(&dd->count, 0); | |
502 | list_add(&dd->list, &t->devices); | |
503 | ||
82b1519b | 504 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 505 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
506 | if (r) |
507 | return r; | |
508 | } | |
509 | atomic_inc(&dd->count); | |
510 | ||
82b1519b | 511 | *result = &dd->dm_dev; |
1da177e4 LT |
512 | return 0; |
513 | } | |
08649012 | 514 | EXPORT_SYMBOL(dm_get_device); |
1da177e4 | 515 | |
754c5fc7 | 516 | int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, |
5dea271b | 517 | sector_t start, sector_t len, void *data) |
1da177e4 | 518 | { |
754c5fc7 MS |
519 | struct queue_limits *limits = data; |
520 | struct block_device *bdev = dev->bdev; | |
165125e1 | 521 | struct request_queue *q = bdev_get_queue(bdev); |
0c2322e4 AK |
522 | char b[BDEVNAME_SIZE]; |
523 | ||
524 | if (unlikely(!q)) { | |
525 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
526 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
754c5fc7 | 527 | return 0; |
0c2322e4 | 528 | } |
3cb40214 | 529 | |
b27d7f16 MP |
530 | if (bdev_stack_limits(limits, bdev, start) < 0) |
531 | DMWARN("%s: adding target device %s caused an alignment inconsistency: " | |
a963a956 MS |
532 | "physical_block_size=%u, logical_block_size=%u, " |
533 | "alignment_offset=%u, start=%llu", | |
534 | dm_device_name(ti->table->md), bdevname(bdev, b), | |
535 | q->limits.physical_block_size, | |
536 | q->limits.logical_block_size, | |
537 | q->limits.alignment_offset, | |
b27d7f16 | 538 | (unsigned long long) start << SECTOR_SHIFT); |
3cb40214 | 539 | |
9980c638 MB |
540 | /* |
541 | * Check if merge fn is supported. | |
542 | * If not we'll force DM to use PAGE_SIZE or | |
543 | * smaller I/O, just to be safe. | |
3cb40214 | 544 | */ |
d5b9dd04 | 545 | if (dm_queue_merge_is_compulsory(q) && !ti->type->merge) |
72d4cd9f MS |
546 | blk_limits_max_hw_sectors(limits, |
547 | (unsigned int) (PAGE_SIZE >> 9)); | |
754c5fc7 | 548 | return 0; |
3cb40214 BR |
549 | } |
550 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 551 | |
1da177e4 | 552 | /* |
08649012 | 553 | * Decrement a device's use count and remove it if necessary. |
1da177e4 | 554 | */ |
82b1519b | 555 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 556 | { |
82b1519b MP |
557 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
558 | dm_dev); | |
559 | ||
1da177e4 | 560 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 561 | close_dev(dd, ti->table->md); |
1da177e4 LT |
562 | list_del(&dd->list); |
563 | kfree(dd); | |
564 | } | |
565 | } | |
08649012 | 566 | EXPORT_SYMBOL(dm_put_device); |
1da177e4 LT |
567 | |
568 | /* | |
569 | * Checks to see if the target joins onto the end of the table. | |
570 | */ | |
571 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
572 | { | |
573 | struct dm_target *prev; | |
574 | ||
575 | if (!table->num_targets) | |
576 | return !ti->begin; | |
577 | ||
578 | prev = &table->targets[table->num_targets - 1]; | |
579 | return (ti->begin == (prev->begin + prev->len)); | |
580 | } | |
581 | ||
582 | /* | |
583 | * Used to dynamically allocate the arg array. | |
584 | */ | |
585 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
586 | { | |
587 | char **argv; | |
588 | unsigned new_size; | |
589 | ||
590 | new_size = *array_size ? *array_size * 2 : 64; | |
591 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
592 | if (argv) { | |
593 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
594 | *array_size = new_size; | |
595 | } | |
596 | ||
597 | kfree(old_argv); | |
598 | return argv; | |
599 | } | |
600 | ||
601 | /* | |
602 | * Destructively splits up the argument list to pass to ctr. | |
603 | */ | |
604 | int dm_split_args(int *argc, char ***argvp, char *input) | |
605 | { | |
606 | char *start, *end = input, *out, **argv = NULL; | |
607 | unsigned array_size = 0; | |
608 | ||
609 | *argc = 0; | |
814d6862 DT |
610 | |
611 | if (!input) { | |
612 | *argvp = NULL; | |
613 | return 0; | |
614 | } | |
615 | ||
1da177e4 LT |
616 | argv = realloc_argv(&array_size, argv); |
617 | if (!argv) | |
618 | return -ENOMEM; | |
619 | ||
620 | while (1) { | |
1da177e4 | 621 | /* Skip whitespace */ |
e7d2860b | 622 | start = skip_spaces(end); |
1da177e4 LT |
623 | |
624 | if (!*start) | |
625 | break; /* success, we hit the end */ | |
626 | ||
627 | /* 'out' is used to remove any back-quotes */ | |
628 | end = out = start; | |
629 | while (*end) { | |
630 | /* Everything apart from '\0' can be quoted */ | |
631 | if (*end == '\\' && *(end + 1)) { | |
632 | *out++ = *(end + 1); | |
633 | end += 2; | |
634 | continue; | |
635 | } | |
636 | ||
637 | if (isspace(*end)) | |
638 | break; /* end of token */ | |
639 | ||
640 | *out++ = *end++; | |
641 | } | |
642 | ||
643 | /* have we already filled the array ? */ | |
644 | if ((*argc + 1) > array_size) { | |
645 | argv = realloc_argv(&array_size, argv); | |
646 | if (!argv) | |
647 | return -ENOMEM; | |
648 | } | |
649 | ||
650 | /* we know this is whitespace */ | |
651 | if (*end) | |
652 | end++; | |
653 | ||
654 | /* terminate the string and put it in the array */ | |
655 | *out = '\0'; | |
656 | argv[*argc] = start; | |
657 | (*argc)++; | |
658 | } | |
659 | ||
660 | *argvp = argv; | |
661 | return 0; | |
662 | } | |
663 | ||
be6d4305 MS |
664 | /* |
665 | * Impose necessary and sufficient conditions on a devices's table such | |
666 | * that any incoming bio which respects its logical_block_size can be | |
667 | * processed successfully. If it falls across the boundary between | |
668 | * two or more targets, the size of each piece it gets split into must | |
669 | * be compatible with the logical_block_size of the target processing it. | |
670 | */ | |
754c5fc7 MS |
671 | static int validate_hardware_logical_block_alignment(struct dm_table *table, |
672 | struct queue_limits *limits) | |
be6d4305 MS |
673 | { |
674 | /* | |
675 | * This function uses arithmetic modulo the logical_block_size | |
676 | * (in units of 512-byte sectors). | |
677 | */ | |
678 | unsigned short device_logical_block_size_sects = | |
754c5fc7 | 679 | limits->logical_block_size >> SECTOR_SHIFT; |
be6d4305 MS |
680 | |
681 | /* | |
682 | * Offset of the start of the next table entry, mod logical_block_size. | |
683 | */ | |
684 | unsigned short next_target_start = 0; | |
685 | ||
686 | /* | |
687 | * Given an aligned bio that extends beyond the end of a | |
688 | * target, how many sectors must the next target handle? | |
689 | */ | |
690 | unsigned short remaining = 0; | |
691 | ||
692 | struct dm_target *uninitialized_var(ti); | |
754c5fc7 | 693 | struct queue_limits ti_limits; |
be6d4305 MS |
694 | unsigned i = 0; |
695 | ||
696 | /* | |
697 | * Check each entry in the table in turn. | |
698 | */ | |
699 | while (i < dm_table_get_num_targets(table)) { | |
700 | ti = dm_table_get_target(table, i++); | |
701 | ||
754c5fc7 MS |
702 | blk_set_default_limits(&ti_limits); |
703 | ||
704 | /* combine all target devices' limits */ | |
705 | if (ti->type->iterate_devices) | |
706 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
707 | &ti_limits); | |
708 | ||
be6d4305 MS |
709 | /* |
710 | * If the remaining sectors fall entirely within this | |
711 | * table entry are they compatible with its logical_block_size? | |
712 | */ | |
713 | if (remaining < ti->len && | |
754c5fc7 | 714 | remaining & ((ti_limits.logical_block_size >> |
be6d4305 MS |
715 | SECTOR_SHIFT) - 1)) |
716 | break; /* Error */ | |
717 | ||
718 | next_target_start = | |
719 | (unsigned short) ((next_target_start + ti->len) & | |
720 | (device_logical_block_size_sects - 1)); | |
721 | remaining = next_target_start ? | |
722 | device_logical_block_size_sects - next_target_start : 0; | |
723 | } | |
724 | ||
725 | if (remaining) { | |
726 | DMWARN("%s: table line %u (start sect %llu len %llu) " | |
a963a956 | 727 | "not aligned to h/w logical block size %u", |
be6d4305 MS |
728 | dm_device_name(table->md), i, |
729 | (unsigned long long) ti->begin, | |
730 | (unsigned long long) ti->len, | |
754c5fc7 | 731 | limits->logical_block_size); |
be6d4305 MS |
732 | return -EINVAL; |
733 | } | |
734 | ||
735 | return 0; | |
736 | } | |
737 | ||
1da177e4 LT |
738 | int dm_table_add_target(struct dm_table *t, const char *type, |
739 | sector_t start, sector_t len, char *params) | |
740 | { | |
741 | int r = -EINVAL, argc; | |
742 | char **argv; | |
743 | struct dm_target *tgt; | |
744 | ||
3791e2fc AK |
745 | if (t->singleton) { |
746 | DMERR("%s: target type %s must appear alone in table", | |
747 | dm_device_name(t->md), t->targets->type->name); | |
748 | return -EINVAL; | |
749 | } | |
750 | ||
1da177e4 LT |
751 | if ((r = check_space(t))) |
752 | return r; | |
753 | ||
754 | tgt = t->targets + t->num_targets; | |
755 | memset(tgt, 0, sizeof(*tgt)); | |
756 | ||
757 | if (!len) { | |
72d94861 | 758 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
759 | return -EINVAL; |
760 | } | |
761 | ||
762 | tgt->type = dm_get_target_type(type); | |
763 | if (!tgt->type) { | |
72d94861 AK |
764 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
765 | type); | |
1da177e4 LT |
766 | return -EINVAL; |
767 | } | |
768 | ||
3791e2fc AK |
769 | if (dm_target_needs_singleton(tgt->type)) { |
770 | if (t->num_targets) { | |
771 | DMERR("%s: target type %s must appear alone in table", | |
772 | dm_device_name(t->md), type); | |
773 | return -EINVAL; | |
774 | } | |
775 | t->singleton = 1; | |
776 | } | |
777 | ||
cc6cbe14 AK |
778 | if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) { |
779 | DMERR("%s: target type %s may not be included in read-only tables", | |
780 | dm_device_name(t->md), type); | |
781 | return -EINVAL; | |
782 | } | |
783 | ||
36a0456f AK |
784 | if (t->immutable_target_type) { |
785 | if (t->immutable_target_type != tgt->type) { | |
786 | DMERR("%s: immutable target type %s cannot be mixed with other target types", | |
787 | dm_device_name(t->md), t->immutable_target_type->name); | |
788 | return -EINVAL; | |
789 | } | |
790 | } else if (dm_target_is_immutable(tgt->type)) { | |
791 | if (t->num_targets) { | |
792 | DMERR("%s: immutable target type %s cannot be mixed with other target types", | |
793 | dm_device_name(t->md), tgt->type->name); | |
794 | return -EINVAL; | |
795 | } | |
796 | t->immutable_target_type = tgt->type; | |
797 | } | |
798 | ||
1da177e4 LT |
799 | tgt->table = t; |
800 | tgt->begin = start; | |
801 | tgt->len = len; | |
802 | tgt->error = "Unknown error"; | |
803 | ||
804 | /* | |
805 | * Does this target adjoin the previous one ? | |
806 | */ | |
807 | if (!adjoin(t, tgt)) { | |
808 | tgt->error = "Gap in table"; | |
809 | r = -EINVAL; | |
810 | goto bad; | |
811 | } | |
812 | ||
813 | r = dm_split_args(&argc, &argv, params); | |
814 | if (r) { | |
815 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
816 | goto bad; | |
817 | } | |
818 | ||
819 | r = tgt->type->ctr(tgt, argc, argv); | |
820 | kfree(argv); | |
821 | if (r) | |
822 | goto bad; | |
823 | ||
824 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
825 | ||
936688d7 MS |
826 | if (!tgt->num_discard_requests && tgt->discards_supported) |
827 | DMWARN("%s: %s: ignoring discards_supported because num_discard_requests is zero.", | |
828 | dm_device_name(t->md), type); | |
5ae89a87 | 829 | |
1da177e4 LT |
830 | return 0; |
831 | ||
832 | bad: | |
72d94861 | 833 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
834 | dm_put_target_type(tgt->type); |
835 | return r; | |
836 | } | |
837 | ||
498f0103 MS |
838 | /* |
839 | * Target argument parsing helpers. | |
840 | */ | |
841 | static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
842 | unsigned *value, char **error, unsigned grouped) | |
843 | { | |
844 | const char *arg_str = dm_shift_arg(arg_set); | |
845 | ||
846 | if (!arg_str || | |
847 | (sscanf(arg_str, "%u", value) != 1) || | |
848 | (*value < arg->min) || | |
849 | (*value > arg->max) || | |
850 | (grouped && arg_set->argc < *value)) { | |
851 | *error = arg->error; | |
852 | return -EINVAL; | |
853 | } | |
854 | ||
855 | return 0; | |
856 | } | |
857 | ||
858 | int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
859 | unsigned *value, char **error) | |
860 | { | |
861 | return validate_next_arg(arg, arg_set, value, error, 0); | |
862 | } | |
863 | EXPORT_SYMBOL(dm_read_arg); | |
864 | ||
865 | int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set, | |
866 | unsigned *value, char **error) | |
867 | { | |
868 | return validate_next_arg(arg, arg_set, value, error, 1); | |
869 | } | |
870 | EXPORT_SYMBOL(dm_read_arg_group); | |
871 | ||
872 | const char *dm_shift_arg(struct dm_arg_set *as) | |
873 | { | |
874 | char *r; | |
875 | ||
876 | if (as->argc) { | |
877 | as->argc--; | |
878 | r = *as->argv; | |
879 | as->argv++; | |
880 | return r; | |
881 | } | |
882 | ||
883 | return NULL; | |
884 | } | |
885 | EXPORT_SYMBOL(dm_shift_arg); | |
886 | ||
887 | void dm_consume_args(struct dm_arg_set *as, unsigned num_args) | |
888 | { | |
889 | BUG_ON(as->argc < num_args); | |
890 | as->argc -= num_args; | |
891 | as->argv += num_args; | |
892 | } | |
893 | EXPORT_SYMBOL(dm_consume_args); | |
894 | ||
26803b9f | 895 | static int dm_table_set_type(struct dm_table *t) |
e6ee8c0b KU |
896 | { |
897 | unsigned i; | |
898 | unsigned bio_based = 0, request_based = 0; | |
899 | struct dm_target *tgt; | |
900 | struct dm_dev_internal *dd; | |
901 | struct list_head *devices; | |
902 | ||
903 | for (i = 0; i < t->num_targets; i++) { | |
904 | tgt = t->targets + i; | |
905 | if (dm_target_request_based(tgt)) | |
906 | request_based = 1; | |
907 | else | |
908 | bio_based = 1; | |
909 | ||
910 | if (bio_based && request_based) { | |
911 | DMWARN("Inconsistent table: different target types" | |
912 | " can't be mixed up"); | |
913 | return -EINVAL; | |
914 | } | |
915 | } | |
916 | ||
917 | if (bio_based) { | |
918 | /* We must use this table as bio-based */ | |
919 | t->type = DM_TYPE_BIO_BASED; | |
920 | return 0; | |
921 | } | |
922 | ||
923 | BUG_ON(!request_based); /* No targets in this table */ | |
924 | ||
925 | /* Non-request-stackable devices can't be used for request-based dm */ | |
926 | devices = dm_table_get_devices(t); | |
927 | list_for_each_entry(dd, devices, list) { | |
928 | if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) { | |
929 | DMWARN("table load rejected: including" | |
930 | " non-request-stackable devices"); | |
931 | return -EINVAL; | |
932 | } | |
933 | } | |
934 | ||
935 | /* | |
936 | * Request-based dm supports only tables that have a single target now. | |
937 | * To support multiple targets, request splitting support is needed, | |
938 | * and that needs lots of changes in the block-layer. | |
939 | * (e.g. request completion process for partial completion.) | |
940 | */ | |
941 | if (t->num_targets > 1) { | |
942 | DMWARN("Request-based dm doesn't support multiple targets yet"); | |
943 | return -EINVAL; | |
944 | } | |
945 | ||
946 | t->type = DM_TYPE_REQUEST_BASED; | |
947 | ||
948 | return 0; | |
949 | } | |
950 | ||
951 | unsigned dm_table_get_type(struct dm_table *t) | |
952 | { | |
953 | return t->type; | |
954 | } | |
955 | ||
36a0456f AK |
956 | struct target_type *dm_table_get_immutable_target_type(struct dm_table *t) |
957 | { | |
958 | return t->immutable_target_type; | |
959 | } | |
960 | ||
e6ee8c0b KU |
961 | bool dm_table_request_based(struct dm_table *t) |
962 | { | |
963 | return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED; | |
964 | } | |
965 | ||
966 | int dm_table_alloc_md_mempools(struct dm_table *t) | |
967 | { | |
968 | unsigned type = dm_table_get_type(t); | |
969 | ||
970 | if (unlikely(type == DM_TYPE_NONE)) { | |
971 | DMWARN("no table type is set, can't allocate mempools"); | |
972 | return -EINVAL; | |
973 | } | |
974 | ||
a91a2785 | 975 | t->mempools = dm_alloc_md_mempools(type, t->integrity_supported); |
e6ee8c0b KU |
976 | if (!t->mempools) |
977 | return -ENOMEM; | |
978 | ||
979 | return 0; | |
980 | } | |
981 | ||
982 | void dm_table_free_md_mempools(struct dm_table *t) | |
983 | { | |
984 | dm_free_md_mempools(t->mempools); | |
985 | t->mempools = NULL; | |
986 | } | |
987 | ||
988 | struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t) | |
989 | { | |
990 | return t->mempools; | |
991 | } | |
992 | ||
1da177e4 LT |
993 | static int setup_indexes(struct dm_table *t) |
994 | { | |
995 | int i; | |
996 | unsigned int total = 0; | |
997 | sector_t *indexes; | |
998 | ||
999 | /* allocate the space for *all* the indexes */ | |
1000 | for (i = t->depth - 2; i >= 0; i--) { | |
1001 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
1002 | total += t->counts[i]; | |
1003 | } | |
1004 | ||
1005 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
1006 | if (!indexes) | |
1007 | return -ENOMEM; | |
1008 | ||
1009 | /* set up internal nodes, bottom-up */ | |
82d601dc | 1010 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
1011 | t->index[i] = indexes; |
1012 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
1013 | setup_btree_index(i, t); | |
1014 | } | |
1015 | ||
1016 | return 0; | |
1017 | } | |
1018 | ||
1019 | /* | |
1020 | * Builds the btree to index the map. | |
1021 | */ | |
26803b9f | 1022 | static int dm_table_build_index(struct dm_table *t) |
1da177e4 LT |
1023 | { |
1024 | int r = 0; | |
1025 | unsigned int leaf_nodes; | |
1026 | ||
1da177e4 LT |
1027 | /* how many indexes will the btree have ? */ |
1028 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
1029 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
1030 | ||
1031 | /* leaf layer has already been set up */ | |
1032 | t->counts[t->depth - 1] = leaf_nodes; | |
1033 | t->index[t->depth - 1] = t->highs; | |
1034 | ||
1035 | if (t->depth >= 2) | |
1036 | r = setup_indexes(t); | |
1037 | ||
1038 | return r; | |
1039 | } | |
1040 | ||
a63a5cf8 MS |
1041 | /* |
1042 | * Get a disk whose integrity profile reflects the table's profile. | |
1043 | * If %match_all is true, all devices' profiles must match. | |
1044 | * If %match_all is false, all devices must at least have an | |
1045 | * allocated integrity profile; but uninitialized is ok. | |
1046 | * Returns NULL if integrity support was inconsistent or unavailable. | |
1047 | */ | |
1048 | static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t, | |
1049 | bool match_all) | |
1050 | { | |
1051 | struct list_head *devices = dm_table_get_devices(t); | |
1052 | struct dm_dev_internal *dd = NULL; | |
1053 | struct gendisk *prev_disk = NULL, *template_disk = NULL; | |
1054 | ||
1055 | list_for_each_entry(dd, devices, list) { | |
1056 | template_disk = dd->dm_dev.bdev->bd_disk; | |
1057 | if (!blk_get_integrity(template_disk)) | |
1058 | goto no_integrity; | |
1059 | if (!match_all && !blk_integrity_is_initialized(template_disk)) | |
1060 | continue; /* skip uninitialized profiles */ | |
1061 | else if (prev_disk && | |
1062 | blk_integrity_compare(prev_disk, template_disk) < 0) | |
1063 | goto no_integrity; | |
1064 | prev_disk = template_disk; | |
1065 | } | |
1066 | ||
1067 | return template_disk; | |
1068 | ||
1069 | no_integrity: | |
1070 | if (prev_disk) | |
1071 | DMWARN("%s: integrity not set: %s and %s profile mismatch", | |
1072 | dm_device_name(t->md), | |
1073 | prev_disk->disk_name, | |
1074 | template_disk->disk_name); | |
1075 | return NULL; | |
1076 | } | |
1077 | ||
26803b9f WD |
1078 | /* |
1079 | * Register the mapped device for blk_integrity support if | |
a63a5cf8 MS |
1080 | * the underlying devices have an integrity profile. But all devices |
1081 | * may not have matching profiles (checking all devices isn't reliable | |
1082 | * during table load because this table may use other DM device(s) which | |
1083 | * must be resumed before they will have an initialized integity profile). | |
1084 | * Stacked DM devices force a 2 stage integrity profile validation: | |
1085 | * 1 - during load, validate all initialized integrity profiles match | |
1086 | * 2 - during resume, validate all integrity profiles match | |
26803b9f WD |
1087 | */ |
1088 | static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md) | |
1089 | { | |
a63a5cf8 | 1090 | struct gendisk *template_disk = NULL; |
26803b9f | 1091 | |
a63a5cf8 MS |
1092 | template_disk = dm_table_get_integrity_disk(t, false); |
1093 | if (!template_disk) | |
1094 | return 0; | |
26803b9f | 1095 | |
a63a5cf8 MS |
1096 | if (!blk_integrity_is_initialized(dm_disk(md))) { |
1097 | t->integrity_supported = 1; | |
1098 | return blk_integrity_register(dm_disk(md), NULL); | |
1099 | } | |
1100 | ||
1101 | /* | |
1102 | * If DM device already has an initalized integrity | |
1103 | * profile the new profile should not conflict. | |
1104 | */ | |
1105 | if (blk_integrity_is_initialized(template_disk) && | |
1106 | blk_integrity_compare(dm_disk(md), template_disk) < 0) { | |
1107 | DMWARN("%s: conflict with existing integrity profile: " | |
1108 | "%s profile mismatch", | |
1109 | dm_device_name(t->md), | |
1110 | template_disk->disk_name); | |
1111 | return 1; | |
1112 | } | |
1113 | ||
1114 | /* Preserve existing initialized integrity profile */ | |
1115 | t->integrity_supported = 1; | |
26803b9f WD |
1116 | return 0; |
1117 | } | |
1118 | ||
1119 | /* | |
1120 | * Prepares the table for use by building the indices, | |
1121 | * setting the type, and allocating mempools. | |
1122 | */ | |
1123 | int dm_table_complete(struct dm_table *t) | |
1124 | { | |
1125 | int r; | |
1126 | ||
1127 | r = dm_table_set_type(t); | |
1128 | if (r) { | |
1129 | DMERR("unable to set table type"); | |
1130 | return r; | |
1131 | } | |
1132 | ||
1133 | r = dm_table_build_index(t); | |
1134 | if (r) { | |
1135 | DMERR("unable to build btrees"); | |
1136 | return r; | |
1137 | } | |
1138 | ||
1139 | r = dm_table_prealloc_integrity(t, t->md); | |
1140 | if (r) { | |
1141 | DMERR("could not register integrity profile."); | |
1142 | return r; | |
1143 | } | |
1144 | ||
1145 | r = dm_table_alloc_md_mempools(t); | |
1146 | if (r) | |
1147 | DMERR("unable to allocate mempools"); | |
1148 | ||
1149 | return r; | |
1150 | } | |
1151 | ||
48c9c27b | 1152 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
1153 | void dm_table_event_callback(struct dm_table *t, |
1154 | void (*fn)(void *), void *context) | |
1155 | { | |
48c9c27b | 1156 | mutex_lock(&_event_lock); |
1da177e4 LT |
1157 | t->event_fn = fn; |
1158 | t->event_context = context; | |
48c9c27b | 1159 | mutex_unlock(&_event_lock); |
1da177e4 LT |
1160 | } |
1161 | ||
1162 | void dm_table_event(struct dm_table *t) | |
1163 | { | |
1164 | /* | |
1165 | * You can no longer call dm_table_event() from interrupt | |
1166 | * context, use a bottom half instead. | |
1167 | */ | |
1168 | BUG_ON(in_interrupt()); | |
1169 | ||
48c9c27b | 1170 | mutex_lock(&_event_lock); |
1da177e4 LT |
1171 | if (t->event_fn) |
1172 | t->event_fn(t->event_context); | |
48c9c27b | 1173 | mutex_unlock(&_event_lock); |
1da177e4 | 1174 | } |
08649012 | 1175 | EXPORT_SYMBOL(dm_table_event); |
1da177e4 LT |
1176 | |
1177 | sector_t dm_table_get_size(struct dm_table *t) | |
1178 | { | |
1179 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
1180 | } | |
08649012 | 1181 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 LT |
1182 | |
1183 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
1184 | { | |
14353539 | 1185 | if (index >= t->num_targets) |
1da177e4 LT |
1186 | return NULL; |
1187 | ||
1188 | return t->targets + index; | |
1189 | } | |
1190 | ||
1191 | /* | |
1192 | * Search the btree for the correct target. | |
512875bd JN |
1193 | * |
1194 | * Caller should check returned pointer with dm_target_is_valid() | |
1195 | * to trap I/O beyond end of device. | |
1da177e4 LT |
1196 | */ |
1197 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
1198 | { | |
1199 | unsigned int l, n = 0, k = 0; | |
1200 | sector_t *node; | |
1201 | ||
1202 | for (l = 0; l < t->depth; l++) { | |
1203 | n = get_child(n, k); | |
1204 | node = get_node(t, l, n); | |
1205 | ||
1206 | for (k = 0; k < KEYS_PER_NODE; k++) | |
1207 | if (node[k] >= sector) | |
1208 | break; | |
1209 | } | |
1210 | ||
1211 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
1212 | } | |
1213 | ||
754c5fc7 MS |
1214 | /* |
1215 | * Establish the new table's queue_limits and validate them. | |
1216 | */ | |
1217 | int dm_calculate_queue_limits(struct dm_table *table, | |
1218 | struct queue_limits *limits) | |
1219 | { | |
1220 | struct dm_target *uninitialized_var(ti); | |
1221 | struct queue_limits ti_limits; | |
1222 | unsigned i = 0; | |
1223 | ||
1224 | blk_set_default_limits(limits); | |
1225 | ||
1226 | while (i < dm_table_get_num_targets(table)) { | |
1227 | blk_set_default_limits(&ti_limits); | |
1228 | ||
1229 | ti = dm_table_get_target(table, i++); | |
1230 | ||
1231 | if (!ti->type->iterate_devices) | |
1232 | goto combine_limits; | |
1233 | ||
1234 | /* | |
1235 | * Combine queue limits of all the devices this target uses. | |
1236 | */ | |
1237 | ti->type->iterate_devices(ti, dm_set_device_limits, | |
1238 | &ti_limits); | |
1239 | ||
40bea431 MS |
1240 | /* Set I/O hints portion of queue limits */ |
1241 | if (ti->type->io_hints) | |
1242 | ti->type->io_hints(ti, &ti_limits); | |
1243 | ||
754c5fc7 MS |
1244 | /* |
1245 | * Check each device area is consistent with the target's | |
1246 | * overall queue limits. | |
1247 | */ | |
f6a1ed10 MP |
1248 | if (ti->type->iterate_devices(ti, device_area_is_invalid, |
1249 | &ti_limits)) | |
754c5fc7 MS |
1250 | return -EINVAL; |
1251 | ||
1252 | combine_limits: | |
1253 | /* | |
1254 | * Merge this target's queue limits into the overall limits | |
1255 | * for the table. | |
1256 | */ | |
1257 | if (blk_stack_limits(limits, &ti_limits, 0) < 0) | |
b27d7f16 | 1258 | DMWARN("%s: adding target device " |
754c5fc7 | 1259 | "(start sect %llu len %llu) " |
b27d7f16 | 1260 | "caused an alignment inconsistency", |
754c5fc7 MS |
1261 | dm_device_name(table->md), |
1262 | (unsigned long long) ti->begin, | |
1263 | (unsigned long long) ti->len); | |
1264 | } | |
1265 | ||
1266 | return validate_hardware_logical_block_alignment(table, limits); | |
1267 | } | |
1268 | ||
9c47008d MP |
1269 | /* |
1270 | * Set the integrity profile for this device if all devices used have | |
a63a5cf8 MS |
1271 | * matching profiles. We're quite deep in the resume path but still |
1272 | * don't know if all devices (particularly DM devices this device | |
1273 | * may be stacked on) have matching profiles. Even if the profiles | |
1274 | * don't match we have no way to fail (to resume) at this point. | |
9c47008d MP |
1275 | */ |
1276 | static void dm_table_set_integrity(struct dm_table *t) | |
1277 | { | |
a63a5cf8 | 1278 | struct gendisk *template_disk = NULL; |
9c47008d MP |
1279 | |
1280 | if (!blk_get_integrity(dm_disk(t->md))) | |
1281 | return; | |
1282 | ||
a63a5cf8 | 1283 | template_disk = dm_table_get_integrity_disk(t, true); |
876fbba1 MS |
1284 | if (template_disk) |
1285 | blk_integrity_register(dm_disk(t->md), | |
1286 | blk_get_integrity(template_disk)); | |
1287 | else if (blk_integrity_is_initialized(dm_disk(t->md))) | |
a63a5cf8 MS |
1288 | DMWARN("%s: device no longer has a valid integrity profile", |
1289 | dm_device_name(t->md)); | |
876fbba1 MS |
1290 | else |
1291 | DMWARN("%s: unable to establish an integrity profile", | |
1292 | dm_device_name(t->md)); | |
9c47008d MP |
1293 | } |
1294 | ||
ed8b752b MS |
1295 | static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, |
1296 | sector_t start, sector_t len, void *data) | |
1297 | { | |
1298 | unsigned flush = (*(unsigned *)data); | |
1299 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1300 | ||
1301 | return q && (q->flush_flags & flush); | |
1302 | } | |
1303 | ||
1304 | static bool dm_table_supports_flush(struct dm_table *t, unsigned flush) | |
1305 | { | |
1306 | struct dm_target *ti; | |
1307 | unsigned i = 0; | |
1308 | ||
1309 | /* | |
1310 | * Require at least one underlying device to support flushes. | |
1311 | * t->devices includes internal dm devices such as mirror logs | |
1312 | * so we need to use iterate_devices here, which targets | |
1313 | * supporting flushes must provide. | |
1314 | */ | |
1315 | while (i < dm_table_get_num_targets(t)) { | |
1316 | ti = dm_table_get_target(t, i++); | |
1317 | ||
1318 | if (!ti->num_flush_requests) | |
1319 | continue; | |
1320 | ||
1321 | if (ti->type->iterate_devices && | |
1322 | ti->type->iterate_devices(ti, device_flush_capable, &flush)) | |
1323 | return 1; | |
1324 | } | |
1325 | ||
1326 | return 0; | |
1327 | } | |
1328 | ||
983c7db3 MB |
1329 | static bool dm_table_discard_zeroes_data(struct dm_table *t) |
1330 | { | |
1331 | struct dm_target *ti; | |
1332 | unsigned i = 0; | |
1333 | ||
1334 | /* Ensure that all targets supports discard_zeroes_data. */ | |
1335 | while (i < dm_table_get_num_targets(t)) { | |
1336 | ti = dm_table_get_target(t, i++); | |
1337 | ||
1338 | if (ti->discard_zeroes_data_unsupported) | |
1339 | return 0; | |
1340 | } | |
1341 | ||
1342 | return 1; | |
1343 | } | |
1344 | ||
4693c966 MSB |
1345 | static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev, |
1346 | sector_t start, sector_t len, void *data) | |
1347 | { | |
1348 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1349 | ||
1350 | return q && blk_queue_nonrot(q); | |
1351 | } | |
1352 | ||
1353 | static bool dm_table_is_nonrot(struct dm_table *t) | |
1354 | { | |
1355 | struct dm_target *ti; | |
1356 | unsigned i = 0; | |
1357 | ||
1358 | /* Ensure that all underlying device are non-rotational. */ | |
1359 | while (i < dm_table_get_num_targets(t)) { | |
1360 | ti = dm_table_get_target(t, i++); | |
1361 | ||
1362 | if (!ti->type->iterate_devices || | |
1363 | !ti->type->iterate_devices(ti, device_is_nonrot, NULL)) | |
1364 | return 0; | |
1365 | } | |
1366 | ||
1367 | return 1; | |
1368 | } | |
1369 | ||
754c5fc7 MS |
1370 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, |
1371 | struct queue_limits *limits) | |
1da177e4 | 1372 | { |
ed8b752b MS |
1373 | unsigned flush = 0; |
1374 | ||
1da177e4 | 1375 | /* |
1197764e | 1376 | * Copy table's limits to the DM device's request_queue |
1da177e4 | 1377 | */ |
754c5fc7 | 1378 | q->limits = *limits; |
c9a3f6d6 | 1379 | |
5ae89a87 MS |
1380 | if (!dm_table_supports_discards(t)) |
1381 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1382 | else | |
1383 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1384 | ||
ed8b752b MS |
1385 | if (dm_table_supports_flush(t, REQ_FLUSH)) { |
1386 | flush |= REQ_FLUSH; | |
1387 | if (dm_table_supports_flush(t, REQ_FUA)) | |
1388 | flush |= REQ_FUA; | |
1389 | } | |
1390 | blk_queue_flush(q, flush); | |
1391 | ||
983c7db3 MB |
1392 | if (!dm_table_discard_zeroes_data(t)) |
1393 | q->limits.discard_zeroes_data = 0; | |
1394 | ||
4693c966 MSB |
1395 | if (dm_table_is_nonrot(t)) |
1396 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q); | |
1397 | else | |
1398 | queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q); | |
1399 | ||
9c47008d | 1400 | dm_table_set_integrity(t); |
e6ee8c0b KU |
1401 | |
1402 | /* | |
1403 | * QUEUE_FLAG_STACKABLE must be set after all queue settings are | |
1404 | * visible to other CPUs because, once the flag is set, incoming bios | |
1405 | * are processed by request-based dm, which refers to the queue | |
1406 | * settings. | |
1407 | * Until the flag set, bios are passed to bio-based dm and queued to | |
1408 | * md->deferred where queue settings are not needed yet. | |
1409 | * Those bios are passed to request-based dm at the resume time. | |
1410 | */ | |
1411 | smp_mb(); | |
1412 | if (dm_table_request_based(t)) | |
1413 | queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q); | |
1da177e4 LT |
1414 | } |
1415 | ||
1416 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
1417 | { | |
1418 | return t->num_targets; | |
1419 | } | |
1420 | ||
1421 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
1422 | { | |
1423 | return &t->devices; | |
1424 | } | |
1425 | ||
aeb5d727 | 1426 | fmode_t dm_table_get_mode(struct dm_table *t) |
1da177e4 LT |
1427 | { |
1428 | return t->mode; | |
1429 | } | |
08649012 | 1430 | EXPORT_SYMBOL(dm_table_get_mode); |
1da177e4 LT |
1431 | |
1432 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
1433 | { | |
1434 | int i = t->num_targets; | |
1435 | struct dm_target *ti = t->targets; | |
1436 | ||
1437 | while (i--) { | |
1438 | if (postsuspend) { | |
1439 | if (ti->type->postsuspend) | |
1440 | ti->type->postsuspend(ti); | |
1441 | } else if (ti->type->presuspend) | |
1442 | ti->type->presuspend(ti); | |
1443 | ||
1444 | ti++; | |
1445 | } | |
1446 | } | |
1447 | ||
1448 | void dm_table_presuspend_targets(struct dm_table *t) | |
1449 | { | |
cf222b37 AK |
1450 | if (!t) |
1451 | return; | |
1452 | ||
e8488d08 | 1453 | suspend_targets(t, 0); |
1da177e4 LT |
1454 | } |
1455 | ||
1456 | void dm_table_postsuspend_targets(struct dm_table *t) | |
1457 | { | |
cf222b37 AK |
1458 | if (!t) |
1459 | return; | |
1460 | ||
e8488d08 | 1461 | suspend_targets(t, 1); |
1da177e4 LT |
1462 | } |
1463 | ||
8757b776 | 1464 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 1465 | { |
8757b776 MB |
1466 | int i, r = 0; |
1467 | ||
1468 | for (i = 0; i < t->num_targets; i++) { | |
1469 | struct dm_target *ti = t->targets + i; | |
1470 | ||
1471 | if (!ti->type->preresume) | |
1472 | continue; | |
1473 | ||
1474 | r = ti->type->preresume(ti); | |
1475 | if (r) | |
1476 | return r; | |
1477 | } | |
1da177e4 LT |
1478 | |
1479 | for (i = 0; i < t->num_targets; i++) { | |
1480 | struct dm_target *ti = t->targets + i; | |
1481 | ||
1482 | if (ti->type->resume) | |
1483 | ti->type->resume(ti); | |
1484 | } | |
8757b776 MB |
1485 | |
1486 | return 0; | |
1da177e4 LT |
1487 | } |
1488 | ||
9d357b07 N |
1489 | void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb) |
1490 | { | |
1491 | list_add(&cb->list, &t->target_callbacks); | |
1492 | } | |
1493 | EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks); | |
1494 | ||
1da177e4 LT |
1495 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) |
1496 | { | |
82b1519b | 1497 | struct dm_dev_internal *dd; |
afb24528 | 1498 | struct list_head *devices = dm_table_get_devices(t); |
9d357b07 | 1499 | struct dm_target_callbacks *cb; |
1da177e4 LT |
1500 | int r = 0; |
1501 | ||
afb24528 | 1502 | list_for_each_entry(dd, devices, list) { |
82b1519b | 1503 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
1504 | char b[BDEVNAME_SIZE]; |
1505 | ||
1506 | if (likely(q)) | |
1507 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
1508 | else | |
1509 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
1510 | dm_device_name(t->md), | |
1511 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
1512 | } |
1513 | ||
9d357b07 N |
1514 | list_for_each_entry(cb, &t->target_callbacks, list) |
1515 | if (cb->congested_fn) | |
1516 | r |= cb->congested_fn(cb, bdi_bits); | |
1517 | ||
1da177e4 LT |
1518 | return r; |
1519 | } | |
1520 | ||
cec47e3d KU |
1521 | int dm_table_any_busy_target(struct dm_table *t) |
1522 | { | |
1523 | unsigned i; | |
1524 | struct dm_target *ti; | |
1525 | ||
1526 | for (i = 0; i < t->num_targets; i++) { | |
1527 | ti = t->targets + i; | |
1528 | if (ti->type->busy && ti->type->busy(ti)) | |
1529 | return 1; | |
1530 | } | |
1531 | ||
1532 | return 0; | |
1533 | } | |
1534 | ||
1134e5ae MA |
1535 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
1536 | { | |
1134e5ae MA |
1537 | return t->md; |
1538 | } | |
08649012 | 1539 | EXPORT_SYMBOL(dm_table_get_md); |
1134e5ae | 1540 | |
5ae89a87 MS |
1541 | static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev, |
1542 | sector_t start, sector_t len, void *data) | |
1543 | { | |
1544 | struct request_queue *q = bdev_get_queue(dev->bdev); | |
1545 | ||
1546 | return q && blk_queue_discard(q); | |
1547 | } | |
1548 | ||
1549 | bool dm_table_supports_discards(struct dm_table *t) | |
1550 | { | |
1551 | struct dm_target *ti; | |
1552 | unsigned i = 0; | |
1553 | ||
5ae89a87 | 1554 | /* |
4c259327 MS |
1555 | * Unless any target used by the table set discards_supported, |
1556 | * require at least one underlying device to support discards. | |
5ae89a87 MS |
1557 | * t->devices includes internal dm devices such as mirror logs |
1558 | * so we need to use iterate_devices here, which targets | |
936688d7 | 1559 | * supporting discard selectively must provide. |
5ae89a87 MS |
1560 | */ |
1561 | while (i < dm_table_get_num_targets(t)) { | |
1562 | ti = dm_table_get_target(t, i++); | |
1563 | ||
936688d7 MS |
1564 | if (!ti->num_discard_requests) |
1565 | continue; | |
1566 | ||
4c259327 MS |
1567 | if (ti->discards_supported) |
1568 | return 1; | |
1569 | ||
5ae89a87 MS |
1570 | if (ti->type->iterate_devices && |
1571 | ti->type->iterate_devices(ti, device_discard_capable, NULL)) | |
1572 | return 1; | |
1573 | } | |
1574 | ||
1575 | return 0; | |
1576 | } |