]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (C) 2001 Sistina Software (UK) Limited. | |
3 | * Copyright (C) 2004 Red Hat, Inc. All rights reserved. | |
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> | |
15 | #include <linux/slab.h> | |
16 | #include <linux/interrupt.h> | |
48c9c27b | 17 | #include <linux/mutex.h> |
1da177e4 LT |
18 | #include <asm/atomic.h> |
19 | ||
72d94861 AK |
20 | #define DM_MSG_PREFIX "table" |
21 | ||
1da177e4 LT |
22 | #define MAX_DEPTH 16 |
23 | #define NODE_SIZE L1_CACHE_BYTES | |
24 | #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t)) | |
25 | #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1) | |
26 | ||
27 | struct dm_table { | |
1134e5ae | 28 | struct mapped_device *md; |
1da177e4 LT |
29 | atomic_t holders; |
30 | ||
31 | /* btree table */ | |
32 | unsigned int depth; | |
33 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | |
34 | sector_t *index[MAX_DEPTH]; | |
35 | ||
36 | unsigned int num_targets; | |
37 | unsigned int num_allocated; | |
38 | sector_t *highs; | |
39 | struct dm_target *targets; | |
40 | ||
41 | /* | |
42 | * Indicates the rw permissions for the new logical | |
43 | * device. This should be a combination of FMODE_READ | |
44 | * and FMODE_WRITE. | |
45 | */ | |
46 | int mode; | |
47 | ||
48 | /* a list of devices used by this table */ | |
49 | struct list_head devices; | |
50 | ||
51 | /* | |
52 | * These are optimistic limits taken from all the | |
53 | * targets, some targets will need smaller limits. | |
54 | */ | |
55 | struct io_restrictions limits; | |
56 | ||
57 | /* events get handed up using this callback */ | |
58 | void (*event_fn)(void *); | |
59 | void *event_context; | |
60 | }; | |
61 | ||
62 | /* | |
63 | * Similar to ceiling(log_size(n)) | |
64 | */ | |
65 | static unsigned int int_log(unsigned int n, unsigned int base) | |
66 | { | |
67 | int result = 0; | |
68 | ||
69 | while (n > 1) { | |
70 | n = dm_div_up(n, base); | |
71 | result++; | |
72 | } | |
73 | ||
74 | return result; | |
75 | } | |
76 | ||
77 | /* | |
78 | * Returns the minimum that is _not_ zero, unless both are zero. | |
79 | */ | |
80 | #define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r)) | |
81 | ||
82 | /* | |
83 | * Combine two io_restrictions, always taking the lower value. | |
84 | */ | |
85 | static void combine_restrictions_low(struct io_restrictions *lhs, | |
86 | struct io_restrictions *rhs) | |
87 | { | |
88 | lhs->max_sectors = | |
89 | min_not_zero(lhs->max_sectors, rhs->max_sectors); | |
90 | ||
91 | lhs->max_phys_segments = | |
92 | min_not_zero(lhs->max_phys_segments, rhs->max_phys_segments); | |
93 | ||
94 | lhs->max_hw_segments = | |
95 | min_not_zero(lhs->max_hw_segments, rhs->max_hw_segments); | |
96 | ||
97 | lhs->hardsect_size = max(lhs->hardsect_size, rhs->hardsect_size); | |
98 | ||
99 | lhs->max_segment_size = | |
100 | min_not_zero(lhs->max_segment_size, rhs->max_segment_size); | |
101 | ||
91212507 NB |
102 | lhs->max_hw_sectors = |
103 | min_not_zero(lhs->max_hw_sectors, rhs->max_hw_sectors); | |
104 | ||
1da177e4 LT |
105 | lhs->seg_boundary_mask = |
106 | min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask); | |
969429b5 | 107 | |
5ec140e6 VA |
108 | lhs->bounce_pfn = min_not_zero(lhs->bounce_pfn, rhs->bounce_pfn); |
109 | ||
969429b5 | 110 | lhs->no_cluster |= rhs->no_cluster; |
1da177e4 LT |
111 | } |
112 | ||
113 | /* | |
114 | * Calculate the index of the child node of the n'th node k'th key. | |
115 | */ | |
116 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
117 | { | |
118 | return (n * CHILDREN_PER_NODE) + k; | |
119 | } | |
120 | ||
121 | /* | |
122 | * Return the n'th node of level l from table t. | |
123 | */ | |
124 | static inline sector_t *get_node(struct dm_table *t, | |
125 | unsigned int l, unsigned int n) | |
126 | { | |
127 | return t->index[l] + (n * KEYS_PER_NODE); | |
128 | } | |
129 | ||
130 | /* | |
131 | * Return the highest key that you could lookup from the n'th | |
132 | * node on level l of the btree. | |
133 | */ | |
134 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
135 | { | |
136 | for (; l < t->depth - 1; l++) | |
137 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
138 | ||
139 | if (n >= t->counts[l]) | |
140 | return (sector_t) - 1; | |
141 | ||
142 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
143 | } | |
144 | ||
145 | /* | |
146 | * Fills in a level of the btree based on the highs of the level | |
147 | * below it. | |
148 | */ | |
149 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
150 | { | |
151 | unsigned int n, k; | |
152 | sector_t *node; | |
153 | ||
154 | for (n = 0U; n < t->counts[l]; n++) { | |
155 | node = get_node(t, l, n); | |
156 | ||
157 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
158 | node[k] = high(t, l + 1, get_child(n, k)); | |
159 | } | |
160 | ||
161 | return 0; | |
162 | } | |
163 | ||
164 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
165 | { | |
166 | unsigned long size; | |
167 | void *addr; | |
168 | ||
169 | /* | |
170 | * Check that we're not going to overflow. | |
171 | */ | |
172 | if (nmemb > (ULONG_MAX / elem_size)) | |
173 | return NULL; | |
174 | ||
175 | size = nmemb * elem_size; | |
176 | addr = vmalloc(size); | |
177 | if (addr) | |
178 | memset(addr, 0, size); | |
179 | ||
180 | return addr; | |
181 | } | |
182 | ||
183 | /* | |
184 | * highs, and targets are managed as dynamic arrays during a | |
185 | * table load. | |
186 | */ | |
187 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
188 | { | |
189 | sector_t *n_highs; | |
190 | struct dm_target *n_targets; | |
191 | int n = t->num_targets; | |
192 | ||
193 | /* | |
194 | * Allocate both the target array and offset array at once. | |
512875bd JN |
195 | * Append an empty entry to catch sectors beyond the end of |
196 | * the device. | |
1da177e4 | 197 | */ |
512875bd | 198 | n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) + |
1da177e4 LT |
199 | sizeof(sector_t)); |
200 | if (!n_highs) | |
201 | return -ENOMEM; | |
202 | ||
203 | n_targets = (struct dm_target *) (n_highs + num); | |
204 | ||
205 | if (n) { | |
206 | memcpy(n_highs, t->highs, sizeof(*n_highs) * n); | |
207 | memcpy(n_targets, t->targets, sizeof(*n_targets) * n); | |
208 | } | |
209 | ||
210 | memset(n_highs + n, -1, sizeof(*n_highs) * (num - n)); | |
211 | vfree(t->highs); | |
212 | ||
213 | t->num_allocated = num; | |
214 | t->highs = n_highs; | |
215 | t->targets = n_targets; | |
216 | ||
217 | return 0; | |
218 | } | |
219 | ||
1134e5ae MA |
220 | int dm_table_create(struct dm_table **result, int mode, |
221 | unsigned num_targets, struct mapped_device *md) | |
1da177e4 | 222 | { |
094262db | 223 | struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL); |
1da177e4 LT |
224 | |
225 | if (!t) | |
226 | return -ENOMEM; | |
227 | ||
1da177e4 LT |
228 | INIT_LIST_HEAD(&t->devices); |
229 | atomic_set(&t->holders, 1); | |
230 | ||
231 | if (!num_targets) | |
232 | num_targets = KEYS_PER_NODE; | |
233 | ||
234 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
235 | ||
236 | if (alloc_targets(t, num_targets)) { | |
237 | kfree(t); | |
238 | t = NULL; | |
239 | return -ENOMEM; | |
240 | } | |
241 | ||
242 | t->mode = mode; | |
1134e5ae | 243 | t->md = md; |
1da177e4 LT |
244 | *result = t; |
245 | return 0; | |
246 | } | |
247 | ||
248 | static void free_devices(struct list_head *devices) | |
249 | { | |
250 | struct list_head *tmp, *next; | |
251 | ||
afb24528 | 252 | list_for_each_safe(tmp, next, devices) { |
82b1519b MP |
253 | struct dm_dev_internal *dd = |
254 | list_entry(tmp, struct dm_dev_internal, list); | |
1da177e4 LT |
255 | kfree(dd); |
256 | } | |
257 | } | |
258 | ||
5e198d94 | 259 | static void table_destroy(struct dm_table *t) |
1da177e4 LT |
260 | { |
261 | unsigned int i; | |
262 | ||
263 | /* free the indexes (see dm_table_complete) */ | |
264 | if (t->depth >= 2) | |
265 | vfree(t->index[t->depth - 2]); | |
266 | ||
267 | /* free the targets */ | |
268 | for (i = 0; i < t->num_targets; i++) { | |
269 | struct dm_target *tgt = t->targets + i; | |
270 | ||
271 | if (tgt->type->dtr) | |
272 | tgt->type->dtr(tgt); | |
273 | ||
274 | dm_put_target_type(tgt->type); | |
275 | } | |
276 | ||
277 | vfree(t->highs); | |
278 | ||
279 | /* free the device list */ | |
280 | if (t->devices.next != &t->devices) { | |
281 | DMWARN("devices still present during destroy: " | |
282 | "dm_table_remove_device calls missing"); | |
283 | ||
284 | free_devices(&t->devices); | |
285 | } | |
286 | ||
287 | kfree(t); | |
288 | } | |
289 | ||
290 | void dm_table_get(struct dm_table *t) | |
291 | { | |
292 | atomic_inc(&t->holders); | |
293 | } | |
294 | ||
295 | void dm_table_put(struct dm_table *t) | |
296 | { | |
297 | if (!t) | |
298 | return; | |
299 | ||
300 | if (atomic_dec_and_test(&t->holders)) | |
301 | table_destroy(t); | |
302 | } | |
303 | ||
304 | /* | |
305 | * Checks to see if we need to extend highs or targets. | |
306 | */ | |
307 | static inline int check_space(struct dm_table *t) | |
308 | { | |
309 | if (t->num_targets >= t->num_allocated) | |
310 | return alloc_targets(t, t->num_allocated * 2); | |
311 | ||
312 | return 0; | |
313 | } | |
314 | ||
315 | /* | |
316 | * Convert a device path to a dev_t. | |
317 | */ | |
318 | static int lookup_device(const char *path, dev_t *dev) | |
319 | { | |
d5686b44 AV |
320 | struct block_device *bdev = lookup_bdev(path); |
321 | if (IS_ERR(bdev)) | |
322 | return PTR_ERR(bdev); | |
323 | *dev = bdev->bd_dev; | |
324 | bdput(bdev); | |
325 | return 0; | |
1da177e4 LT |
326 | } |
327 | ||
328 | /* | |
329 | * See if we've already got a device in the list. | |
330 | */ | |
82b1519b | 331 | static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev) |
1da177e4 | 332 | { |
82b1519b | 333 | struct dm_dev_internal *dd; |
1da177e4 LT |
334 | |
335 | list_for_each_entry (dd, l, list) | |
82b1519b | 336 | if (dd->dm_dev.bdev->bd_dev == dev) |
1da177e4 LT |
337 | return dd; |
338 | ||
339 | return NULL; | |
340 | } | |
341 | ||
342 | /* | |
343 | * Open a device so we can use it as a map destination. | |
344 | */ | |
82b1519b MP |
345 | static int open_dev(struct dm_dev_internal *d, dev_t dev, |
346 | struct mapped_device *md) | |
1da177e4 LT |
347 | { |
348 | static char *_claim_ptr = "I belong to device-mapper"; | |
349 | struct block_device *bdev; | |
350 | ||
351 | int r; | |
352 | ||
82b1519b | 353 | BUG_ON(d->dm_dev.bdev); |
1da177e4 | 354 | |
82b1519b | 355 | bdev = open_by_devnum(dev, d->dm_dev.mode); |
1da177e4 LT |
356 | if (IS_ERR(bdev)) |
357 | return PTR_ERR(bdev); | |
f165921d | 358 | r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md)); |
1da177e4 LT |
359 | if (r) |
360 | blkdev_put(bdev); | |
361 | else | |
82b1519b | 362 | d->dm_dev.bdev = bdev; |
1da177e4 LT |
363 | return r; |
364 | } | |
365 | ||
366 | /* | |
367 | * Close a device that we've been using. | |
368 | */ | |
82b1519b | 369 | static void close_dev(struct dm_dev_internal *d, struct mapped_device *md) |
1da177e4 | 370 | { |
82b1519b | 371 | if (!d->dm_dev.bdev) |
1da177e4 LT |
372 | return; |
373 | ||
82b1519b MP |
374 | bd_release_from_disk(d->dm_dev.bdev, dm_disk(md)); |
375 | blkdev_put(d->dm_dev.bdev); | |
376 | d->dm_dev.bdev = NULL; | |
1da177e4 LT |
377 | } |
378 | ||
379 | /* | |
2cd54d9b | 380 | * If possible, this checks an area of a destination device is valid. |
1da177e4 | 381 | */ |
82b1519b MP |
382 | static int check_device_area(struct dm_dev_internal *dd, sector_t start, |
383 | sector_t len) | |
1da177e4 | 384 | { |
82b1519b | 385 | sector_t dev_size = dd->dm_dev.bdev->bd_inode->i_size >> SECTOR_SHIFT; |
2cd54d9b MA |
386 | |
387 | if (!dev_size) | |
388 | return 1; | |
389 | ||
1da177e4 LT |
390 | return ((start < dev_size) && (len <= (dev_size - start))); |
391 | } | |
392 | ||
393 | /* | |
394 | * This upgrades the mode on an already open dm_dev. Being | |
395 | * careful to leave things as they were if we fail to reopen the | |
396 | * device. | |
397 | */ | |
82b1519b MP |
398 | static int upgrade_mode(struct dm_dev_internal *dd, int new_mode, |
399 | struct mapped_device *md) | |
1da177e4 LT |
400 | { |
401 | int r; | |
82b1519b MP |
402 | struct dm_dev_internal dd_copy; |
403 | dev_t dev = dd->dm_dev.bdev->bd_dev; | |
1da177e4 LT |
404 | |
405 | dd_copy = *dd; | |
406 | ||
82b1519b MP |
407 | dd->dm_dev.mode |= new_mode; |
408 | dd->dm_dev.bdev = NULL; | |
f165921d | 409 | r = open_dev(dd, dev, md); |
1da177e4 | 410 | if (!r) |
f165921d | 411 | close_dev(&dd_copy, md); |
1da177e4 LT |
412 | else |
413 | *dd = dd_copy; | |
414 | ||
415 | return r; | |
416 | } | |
417 | ||
418 | /* | |
419 | * Add a device to the list, or just increment the usage count if | |
420 | * it's already present. | |
421 | */ | |
422 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
423 | const char *path, sector_t start, sector_t len, | |
424 | int mode, struct dm_dev **result) | |
425 | { | |
426 | int r; | |
69a2ce72 | 427 | dev_t uninitialized_var(dev); |
82b1519b | 428 | struct dm_dev_internal *dd; |
1da177e4 LT |
429 | unsigned int major, minor; |
430 | ||
547bc926 | 431 | BUG_ON(!t); |
1da177e4 LT |
432 | |
433 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
434 | /* Extract the major/minor numbers */ | |
435 | dev = MKDEV(major, minor); | |
436 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
437 | return -EOVERFLOW; | |
438 | } else { | |
439 | /* convert the path to a device */ | |
440 | if ((r = lookup_device(path, &dev))) | |
441 | return r; | |
442 | } | |
443 | ||
444 | dd = find_device(&t->devices, dev); | |
445 | if (!dd) { | |
446 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
447 | if (!dd) | |
448 | return -ENOMEM; | |
449 | ||
82b1519b MP |
450 | dd->dm_dev.mode = mode; |
451 | dd->dm_dev.bdev = NULL; | |
1da177e4 | 452 | |
f165921d | 453 | if ((r = open_dev(dd, dev, t->md))) { |
1da177e4 LT |
454 | kfree(dd); |
455 | return r; | |
456 | } | |
457 | ||
82b1519b | 458 | format_dev_t(dd->dm_dev.name, dev); |
1da177e4 LT |
459 | |
460 | atomic_set(&dd->count, 0); | |
461 | list_add(&dd->list, &t->devices); | |
462 | ||
82b1519b | 463 | } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) { |
f165921d | 464 | r = upgrade_mode(dd, mode, t->md); |
1da177e4 LT |
465 | if (r) |
466 | return r; | |
467 | } | |
468 | atomic_inc(&dd->count); | |
469 | ||
470 | if (!check_device_area(dd, start, len)) { | |
471 | DMWARN("device %s too small for target", path); | |
82b1519b | 472 | dm_put_device(ti, &dd->dm_dev); |
1da177e4 LT |
473 | return -EINVAL; |
474 | } | |
475 | ||
82b1519b | 476 | *result = &dd->dm_dev; |
1da177e4 LT |
477 | |
478 | return 0; | |
479 | } | |
480 | ||
3cb40214 | 481 | void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev) |
1da177e4 | 482 | { |
165125e1 | 483 | struct request_queue *q = bdev_get_queue(bdev); |
3cb40214 | 484 | struct io_restrictions *rs = &ti->limits; |
0c2322e4 AK |
485 | char b[BDEVNAME_SIZE]; |
486 | ||
487 | if (unlikely(!q)) { | |
488 | DMWARN("%s: Cannot set limits for nonexistent device %s", | |
489 | dm_device_name(ti->table->md), bdevname(bdev, b)); | |
490 | return; | |
491 | } | |
3cb40214 BR |
492 | |
493 | /* | |
494 | * Combine the device limits low. | |
495 | * | |
496 | * FIXME: if we move an io_restriction struct | |
497 | * into q this would just be a call to | |
498 | * combine_restrictions_low() | |
499 | */ | |
500 | rs->max_sectors = | |
501 | min_not_zero(rs->max_sectors, q->max_sectors); | |
502 | ||
9980c638 MB |
503 | /* |
504 | * Check if merge fn is supported. | |
505 | * If not we'll force DM to use PAGE_SIZE or | |
506 | * smaller I/O, just to be safe. | |
3cb40214 | 507 | */ |
9980c638 MB |
508 | |
509 | if (q->merge_bvec_fn && !ti->type->merge) | |
1da177e4 | 510 | rs->max_sectors = |
3cb40214 BR |
511 | min_not_zero(rs->max_sectors, |
512 | (unsigned int) (PAGE_SIZE >> 9)); | |
1da177e4 | 513 | |
3cb40214 BR |
514 | rs->max_phys_segments = |
515 | min_not_zero(rs->max_phys_segments, | |
516 | q->max_phys_segments); | |
1da177e4 | 517 | |
3cb40214 BR |
518 | rs->max_hw_segments = |
519 | min_not_zero(rs->max_hw_segments, q->max_hw_segments); | |
1da177e4 | 520 | |
3cb40214 | 521 | rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size); |
1da177e4 | 522 | |
3cb40214 BR |
523 | rs->max_segment_size = |
524 | min_not_zero(rs->max_segment_size, q->max_segment_size); | |
1da177e4 | 525 | |
91212507 NB |
526 | rs->max_hw_sectors = |
527 | min_not_zero(rs->max_hw_sectors, q->max_hw_sectors); | |
528 | ||
3cb40214 BR |
529 | rs->seg_boundary_mask = |
530 | min_not_zero(rs->seg_boundary_mask, | |
531 | q->seg_boundary_mask); | |
1da177e4 | 532 | |
5ec140e6 VA |
533 | rs->bounce_pfn = min_not_zero(rs->bounce_pfn, q->bounce_pfn); |
534 | ||
3cb40214 BR |
535 | rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags); |
536 | } | |
537 | EXPORT_SYMBOL_GPL(dm_set_device_limits); | |
969429b5 | 538 | |
3cb40214 BR |
539 | int dm_get_device(struct dm_target *ti, const char *path, sector_t start, |
540 | sector_t len, int mode, struct dm_dev **result) | |
541 | { | |
542 | int r = __table_get_device(ti->table, ti, path, | |
543 | start, len, mode, result); | |
544 | ||
545 | if (!r) | |
546 | dm_set_device_limits(ti, (*result)->bdev); | |
1da177e4 LT |
547 | |
548 | return r; | |
549 | } | |
550 | ||
551 | /* | |
552 | * Decrement a devices use count and remove it if necessary. | |
553 | */ | |
82b1519b | 554 | void dm_put_device(struct dm_target *ti, struct dm_dev *d) |
1da177e4 | 555 | { |
82b1519b MP |
556 | struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal, |
557 | dm_dev); | |
558 | ||
1da177e4 | 559 | if (atomic_dec_and_test(&dd->count)) { |
f165921d | 560 | close_dev(dd, ti->table->md); |
1da177e4 LT |
561 | list_del(&dd->list); |
562 | kfree(dd); | |
563 | } | |
564 | } | |
565 | ||
566 | /* | |
567 | * Checks to see if the target joins onto the end of the table. | |
568 | */ | |
569 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
570 | { | |
571 | struct dm_target *prev; | |
572 | ||
573 | if (!table->num_targets) | |
574 | return !ti->begin; | |
575 | ||
576 | prev = &table->targets[table->num_targets - 1]; | |
577 | return (ti->begin == (prev->begin + prev->len)); | |
578 | } | |
579 | ||
580 | /* | |
581 | * Used to dynamically allocate the arg array. | |
582 | */ | |
583 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
584 | { | |
585 | char **argv; | |
586 | unsigned new_size; | |
587 | ||
588 | new_size = *array_size ? *array_size * 2 : 64; | |
589 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
590 | if (argv) { | |
591 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
592 | *array_size = new_size; | |
593 | } | |
594 | ||
595 | kfree(old_argv); | |
596 | return argv; | |
597 | } | |
598 | ||
599 | /* | |
600 | * Destructively splits up the argument list to pass to ctr. | |
601 | */ | |
602 | int dm_split_args(int *argc, char ***argvp, char *input) | |
603 | { | |
604 | char *start, *end = input, *out, **argv = NULL; | |
605 | unsigned array_size = 0; | |
606 | ||
607 | *argc = 0; | |
814d6862 DT |
608 | |
609 | if (!input) { | |
610 | *argvp = NULL; | |
611 | return 0; | |
612 | } | |
613 | ||
1da177e4 LT |
614 | argv = realloc_argv(&array_size, argv); |
615 | if (!argv) | |
616 | return -ENOMEM; | |
617 | ||
618 | while (1) { | |
619 | start = end; | |
620 | ||
621 | /* Skip whitespace */ | |
622 | while (*start && isspace(*start)) | |
623 | start++; | |
624 | ||
625 | if (!*start) | |
626 | break; /* success, we hit the end */ | |
627 | ||
628 | /* 'out' is used to remove any back-quotes */ | |
629 | end = out = start; | |
630 | while (*end) { | |
631 | /* Everything apart from '\0' can be quoted */ | |
632 | if (*end == '\\' && *(end + 1)) { | |
633 | *out++ = *(end + 1); | |
634 | end += 2; | |
635 | continue; | |
636 | } | |
637 | ||
638 | if (isspace(*end)) | |
639 | break; /* end of token */ | |
640 | ||
641 | *out++ = *end++; | |
642 | } | |
643 | ||
644 | /* have we already filled the array ? */ | |
645 | if ((*argc + 1) > array_size) { | |
646 | argv = realloc_argv(&array_size, argv); | |
647 | if (!argv) | |
648 | return -ENOMEM; | |
649 | } | |
650 | ||
651 | /* we know this is whitespace */ | |
652 | if (*end) | |
653 | end++; | |
654 | ||
655 | /* terminate the string and put it in the array */ | |
656 | *out = '\0'; | |
657 | argv[*argc] = start; | |
658 | (*argc)++; | |
659 | } | |
660 | ||
661 | *argvp = argv; | |
662 | return 0; | |
663 | } | |
664 | ||
665 | static void check_for_valid_limits(struct io_restrictions *rs) | |
666 | { | |
667 | if (!rs->max_sectors) | |
defd94b7 | 668 | rs->max_sectors = SAFE_MAX_SECTORS; |
91212507 NB |
669 | if (!rs->max_hw_sectors) |
670 | rs->max_hw_sectors = SAFE_MAX_SECTORS; | |
1da177e4 LT |
671 | if (!rs->max_phys_segments) |
672 | rs->max_phys_segments = MAX_PHYS_SEGMENTS; | |
673 | if (!rs->max_hw_segments) | |
674 | rs->max_hw_segments = MAX_HW_SEGMENTS; | |
675 | if (!rs->hardsect_size) | |
676 | rs->hardsect_size = 1 << SECTOR_SHIFT; | |
677 | if (!rs->max_segment_size) | |
678 | rs->max_segment_size = MAX_SEGMENT_SIZE; | |
679 | if (!rs->seg_boundary_mask) | |
680 | rs->seg_boundary_mask = -1; | |
5ec140e6 VA |
681 | if (!rs->bounce_pfn) |
682 | rs->bounce_pfn = -1; | |
1da177e4 LT |
683 | } |
684 | ||
685 | int dm_table_add_target(struct dm_table *t, const char *type, | |
686 | sector_t start, sector_t len, char *params) | |
687 | { | |
688 | int r = -EINVAL, argc; | |
689 | char **argv; | |
690 | struct dm_target *tgt; | |
691 | ||
692 | if ((r = check_space(t))) | |
693 | return r; | |
694 | ||
695 | tgt = t->targets + t->num_targets; | |
696 | memset(tgt, 0, sizeof(*tgt)); | |
697 | ||
698 | if (!len) { | |
72d94861 | 699 | DMERR("%s: zero-length target", dm_device_name(t->md)); |
1da177e4 LT |
700 | return -EINVAL; |
701 | } | |
702 | ||
703 | tgt->type = dm_get_target_type(type); | |
704 | if (!tgt->type) { | |
72d94861 AK |
705 | DMERR("%s: %s: unknown target type", dm_device_name(t->md), |
706 | type); | |
1da177e4 LT |
707 | return -EINVAL; |
708 | } | |
709 | ||
710 | tgt->table = t; | |
711 | tgt->begin = start; | |
712 | tgt->len = len; | |
713 | tgt->error = "Unknown error"; | |
714 | ||
715 | /* | |
716 | * Does this target adjoin the previous one ? | |
717 | */ | |
718 | if (!adjoin(t, tgt)) { | |
719 | tgt->error = "Gap in table"; | |
720 | r = -EINVAL; | |
721 | goto bad; | |
722 | } | |
723 | ||
724 | r = dm_split_args(&argc, &argv, params); | |
725 | if (r) { | |
726 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
727 | goto bad; | |
728 | } | |
729 | ||
730 | r = tgt->type->ctr(tgt, argc, argv); | |
731 | kfree(argv); | |
732 | if (r) | |
733 | goto bad; | |
734 | ||
735 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
736 | ||
737 | /* FIXME: the plan is to combine high here and then have | |
738 | * the merge fn apply the target level restrictions. */ | |
739 | combine_restrictions_low(&t->limits, &tgt->limits); | |
740 | return 0; | |
741 | ||
742 | bad: | |
72d94861 | 743 | DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error); |
1da177e4 LT |
744 | dm_put_target_type(tgt->type); |
745 | return r; | |
746 | } | |
747 | ||
748 | static int setup_indexes(struct dm_table *t) | |
749 | { | |
750 | int i; | |
751 | unsigned int total = 0; | |
752 | sector_t *indexes; | |
753 | ||
754 | /* allocate the space for *all* the indexes */ | |
755 | for (i = t->depth - 2; i >= 0; i--) { | |
756 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
757 | total += t->counts[i]; | |
758 | } | |
759 | ||
760 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
761 | if (!indexes) | |
762 | return -ENOMEM; | |
763 | ||
764 | /* set up internal nodes, bottom-up */ | |
82d601dc | 765 | for (i = t->depth - 2; i >= 0; i--) { |
1da177e4 LT |
766 | t->index[i] = indexes; |
767 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
768 | setup_btree_index(i, t); | |
769 | } | |
770 | ||
771 | return 0; | |
772 | } | |
773 | ||
774 | /* | |
775 | * Builds the btree to index the map. | |
776 | */ | |
777 | int dm_table_complete(struct dm_table *t) | |
778 | { | |
779 | int r = 0; | |
780 | unsigned int leaf_nodes; | |
781 | ||
782 | check_for_valid_limits(&t->limits); | |
783 | ||
784 | /* how many indexes will the btree have ? */ | |
785 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
786 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
787 | ||
788 | /* leaf layer has already been set up */ | |
789 | t->counts[t->depth - 1] = leaf_nodes; | |
790 | t->index[t->depth - 1] = t->highs; | |
791 | ||
792 | if (t->depth >= 2) | |
793 | r = setup_indexes(t); | |
794 | ||
795 | return r; | |
796 | } | |
797 | ||
48c9c27b | 798 | static DEFINE_MUTEX(_event_lock); |
1da177e4 LT |
799 | void dm_table_event_callback(struct dm_table *t, |
800 | void (*fn)(void *), void *context) | |
801 | { | |
48c9c27b | 802 | mutex_lock(&_event_lock); |
1da177e4 LT |
803 | t->event_fn = fn; |
804 | t->event_context = context; | |
48c9c27b | 805 | mutex_unlock(&_event_lock); |
1da177e4 LT |
806 | } |
807 | ||
808 | void dm_table_event(struct dm_table *t) | |
809 | { | |
810 | /* | |
811 | * You can no longer call dm_table_event() from interrupt | |
812 | * context, use a bottom half instead. | |
813 | */ | |
814 | BUG_ON(in_interrupt()); | |
815 | ||
48c9c27b | 816 | mutex_lock(&_event_lock); |
1da177e4 LT |
817 | if (t->event_fn) |
818 | t->event_fn(t->event_context); | |
48c9c27b | 819 | mutex_unlock(&_event_lock); |
1da177e4 LT |
820 | } |
821 | ||
822 | sector_t dm_table_get_size(struct dm_table *t) | |
823 | { | |
824 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
825 | } | |
826 | ||
827 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
828 | { | |
14353539 | 829 | if (index >= t->num_targets) |
1da177e4 LT |
830 | return NULL; |
831 | ||
832 | return t->targets + index; | |
833 | } | |
834 | ||
835 | /* | |
836 | * Search the btree for the correct target. | |
512875bd JN |
837 | * |
838 | * Caller should check returned pointer with dm_target_is_valid() | |
839 | * to trap I/O beyond end of device. | |
1da177e4 LT |
840 | */ |
841 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
842 | { | |
843 | unsigned int l, n = 0, k = 0; | |
844 | sector_t *node; | |
845 | ||
846 | for (l = 0; l < t->depth; l++) { | |
847 | n = get_child(n, k); | |
848 | node = get_node(t, l, n); | |
849 | ||
850 | for (k = 0; k < KEYS_PER_NODE; k++) | |
851 | if (node[k] >= sector) | |
852 | break; | |
853 | } | |
854 | ||
855 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
856 | } | |
857 | ||
858 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q) | |
859 | { | |
860 | /* | |
861 | * Make sure we obey the optimistic sub devices | |
862 | * restrictions. | |
863 | */ | |
864 | blk_queue_max_sectors(q, t->limits.max_sectors); | |
865 | q->max_phys_segments = t->limits.max_phys_segments; | |
866 | q->max_hw_segments = t->limits.max_hw_segments; | |
867 | q->hardsect_size = t->limits.hardsect_size; | |
868 | q->max_segment_size = t->limits.max_segment_size; | |
91212507 | 869 | q->max_hw_sectors = t->limits.max_hw_sectors; |
1da177e4 | 870 | q->seg_boundary_mask = t->limits.seg_boundary_mask; |
5ec140e6 | 871 | q->bounce_pfn = t->limits.bounce_pfn; |
c9a3f6d6 | 872 | |
969429b5 | 873 | if (t->limits.no_cluster) |
c9a3f6d6 | 874 | queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 875 | else |
c9a3f6d6 | 876 | queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q); |
969429b5 | 877 | |
1da177e4 LT |
878 | } |
879 | ||
880 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
881 | { | |
882 | return t->num_targets; | |
883 | } | |
884 | ||
885 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
886 | { | |
887 | return &t->devices; | |
888 | } | |
889 | ||
890 | int dm_table_get_mode(struct dm_table *t) | |
891 | { | |
892 | return t->mode; | |
893 | } | |
894 | ||
895 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
896 | { | |
897 | int i = t->num_targets; | |
898 | struct dm_target *ti = t->targets; | |
899 | ||
900 | while (i--) { | |
901 | if (postsuspend) { | |
902 | if (ti->type->postsuspend) | |
903 | ti->type->postsuspend(ti); | |
904 | } else if (ti->type->presuspend) | |
905 | ti->type->presuspend(ti); | |
906 | ||
907 | ti++; | |
908 | } | |
909 | } | |
910 | ||
911 | void dm_table_presuspend_targets(struct dm_table *t) | |
912 | { | |
cf222b37 AK |
913 | if (!t) |
914 | return; | |
915 | ||
e8488d08 | 916 | suspend_targets(t, 0); |
1da177e4 LT |
917 | } |
918 | ||
919 | void dm_table_postsuspend_targets(struct dm_table *t) | |
920 | { | |
cf222b37 AK |
921 | if (!t) |
922 | return; | |
923 | ||
e8488d08 | 924 | suspend_targets(t, 1); |
1da177e4 LT |
925 | } |
926 | ||
8757b776 | 927 | int dm_table_resume_targets(struct dm_table *t) |
1da177e4 | 928 | { |
8757b776 MB |
929 | int i, r = 0; |
930 | ||
931 | for (i = 0; i < t->num_targets; i++) { | |
932 | struct dm_target *ti = t->targets + i; | |
933 | ||
934 | if (!ti->type->preresume) | |
935 | continue; | |
936 | ||
937 | r = ti->type->preresume(ti); | |
938 | if (r) | |
939 | return r; | |
940 | } | |
1da177e4 LT |
941 | |
942 | for (i = 0; i < t->num_targets; i++) { | |
943 | struct dm_target *ti = t->targets + i; | |
944 | ||
945 | if (ti->type->resume) | |
946 | ti->type->resume(ti); | |
947 | } | |
8757b776 MB |
948 | |
949 | return 0; | |
1da177e4 LT |
950 | } |
951 | ||
952 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | |
953 | { | |
82b1519b | 954 | struct dm_dev_internal *dd; |
afb24528 | 955 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 LT |
956 | int r = 0; |
957 | ||
afb24528 | 958 | list_for_each_entry(dd, devices, list) { |
82b1519b | 959 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
960 | char b[BDEVNAME_SIZE]; |
961 | ||
962 | if (likely(q)) | |
963 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
964 | else | |
965 | DMWARN_LIMIT("%s: any_congested: nonexistent device %s", | |
966 | dm_device_name(t->md), | |
967 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
968 | } |
969 | ||
970 | return r; | |
971 | } | |
972 | ||
973 | void dm_table_unplug_all(struct dm_table *t) | |
974 | { | |
82b1519b | 975 | struct dm_dev_internal *dd; |
afb24528 | 976 | struct list_head *devices = dm_table_get_devices(t); |
1da177e4 | 977 | |
afb24528 | 978 | list_for_each_entry(dd, devices, list) { |
82b1519b | 979 | struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev); |
0c2322e4 AK |
980 | char b[BDEVNAME_SIZE]; |
981 | ||
982 | if (likely(q)) | |
983 | blk_unplug(q); | |
984 | else | |
985 | DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s", | |
986 | dm_device_name(t->md), | |
987 | bdevname(dd->dm_dev.bdev, b)); | |
1da177e4 LT |
988 | } |
989 | } | |
990 | ||
1134e5ae MA |
991 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
992 | { | |
993 | dm_get(t->md); | |
994 | ||
995 | return t->md; | |
996 | } | |
997 | ||
1da177e4 LT |
998 | EXPORT_SYMBOL(dm_vcalloc); |
999 | EXPORT_SYMBOL(dm_get_device); | |
1000 | EXPORT_SYMBOL(dm_put_device); | |
1001 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 1002 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 1003 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 1004 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
1005 | EXPORT_SYMBOL(dm_table_put); |
1006 | EXPORT_SYMBOL(dm_table_get); | |
1007 | EXPORT_SYMBOL(dm_table_unplug_all); |