]>
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> | |
17 | #include <asm/atomic.h> | |
18 | ||
19 | #define MAX_DEPTH 16 | |
20 | #define NODE_SIZE L1_CACHE_BYTES | |
21 | #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t)) | |
22 | #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1) | |
23 | ||
24 | struct dm_table { | |
1134e5ae | 25 | struct mapped_device *md; |
1da177e4 LT |
26 | atomic_t holders; |
27 | ||
28 | /* btree table */ | |
29 | unsigned int depth; | |
30 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | |
31 | sector_t *index[MAX_DEPTH]; | |
32 | ||
33 | unsigned int num_targets; | |
34 | unsigned int num_allocated; | |
35 | sector_t *highs; | |
36 | struct dm_target *targets; | |
37 | ||
38 | /* | |
39 | * Indicates the rw permissions for the new logical | |
40 | * device. This should be a combination of FMODE_READ | |
41 | * and FMODE_WRITE. | |
42 | */ | |
43 | int mode; | |
44 | ||
45 | /* a list of devices used by this table */ | |
46 | struct list_head devices; | |
47 | ||
48 | /* | |
49 | * These are optimistic limits taken from all the | |
50 | * targets, some targets will need smaller limits. | |
51 | */ | |
52 | struct io_restrictions limits; | |
53 | ||
54 | /* events get handed up using this callback */ | |
55 | void (*event_fn)(void *); | |
56 | void *event_context; | |
57 | }; | |
58 | ||
59 | /* | |
60 | * Similar to ceiling(log_size(n)) | |
61 | */ | |
62 | static unsigned int int_log(unsigned int n, unsigned int base) | |
63 | { | |
64 | int result = 0; | |
65 | ||
66 | while (n > 1) { | |
67 | n = dm_div_up(n, base); | |
68 | result++; | |
69 | } | |
70 | ||
71 | return result; | |
72 | } | |
73 | ||
74 | /* | |
75 | * Returns the minimum that is _not_ zero, unless both are zero. | |
76 | */ | |
77 | #define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r)) | |
78 | ||
79 | /* | |
80 | * Combine two io_restrictions, always taking the lower value. | |
81 | */ | |
82 | static void combine_restrictions_low(struct io_restrictions *lhs, | |
83 | struct io_restrictions *rhs) | |
84 | { | |
85 | lhs->max_sectors = | |
86 | min_not_zero(lhs->max_sectors, rhs->max_sectors); | |
87 | ||
88 | lhs->max_phys_segments = | |
89 | min_not_zero(lhs->max_phys_segments, rhs->max_phys_segments); | |
90 | ||
91 | lhs->max_hw_segments = | |
92 | min_not_zero(lhs->max_hw_segments, rhs->max_hw_segments); | |
93 | ||
94 | lhs->hardsect_size = max(lhs->hardsect_size, rhs->hardsect_size); | |
95 | ||
96 | lhs->max_segment_size = | |
97 | min_not_zero(lhs->max_segment_size, rhs->max_segment_size); | |
98 | ||
99 | lhs->seg_boundary_mask = | |
100 | min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask); | |
969429b5 N |
101 | |
102 | lhs->no_cluster |= rhs->no_cluster; | |
1da177e4 LT |
103 | } |
104 | ||
105 | /* | |
106 | * Calculate the index of the child node of the n'th node k'th key. | |
107 | */ | |
108 | static inline unsigned int get_child(unsigned int n, unsigned int k) | |
109 | { | |
110 | return (n * CHILDREN_PER_NODE) + k; | |
111 | } | |
112 | ||
113 | /* | |
114 | * Return the n'th node of level l from table t. | |
115 | */ | |
116 | static inline sector_t *get_node(struct dm_table *t, | |
117 | unsigned int l, unsigned int n) | |
118 | { | |
119 | return t->index[l] + (n * KEYS_PER_NODE); | |
120 | } | |
121 | ||
122 | /* | |
123 | * Return the highest key that you could lookup from the n'th | |
124 | * node on level l of the btree. | |
125 | */ | |
126 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | |
127 | { | |
128 | for (; l < t->depth - 1; l++) | |
129 | n = get_child(n, CHILDREN_PER_NODE - 1); | |
130 | ||
131 | if (n >= t->counts[l]) | |
132 | return (sector_t) - 1; | |
133 | ||
134 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | |
135 | } | |
136 | ||
137 | /* | |
138 | * Fills in a level of the btree based on the highs of the level | |
139 | * below it. | |
140 | */ | |
141 | static int setup_btree_index(unsigned int l, struct dm_table *t) | |
142 | { | |
143 | unsigned int n, k; | |
144 | sector_t *node; | |
145 | ||
146 | for (n = 0U; n < t->counts[l]; n++) { | |
147 | node = get_node(t, l, n); | |
148 | ||
149 | for (k = 0U; k < KEYS_PER_NODE; k++) | |
150 | node[k] = high(t, l + 1, get_child(n, k)); | |
151 | } | |
152 | ||
153 | return 0; | |
154 | } | |
155 | ||
156 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | |
157 | { | |
158 | unsigned long size; | |
159 | void *addr; | |
160 | ||
161 | /* | |
162 | * Check that we're not going to overflow. | |
163 | */ | |
164 | if (nmemb > (ULONG_MAX / elem_size)) | |
165 | return NULL; | |
166 | ||
167 | size = nmemb * elem_size; | |
168 | addr = vmalloc(size); | |
169 | if (addr) | |
170 | memset(addr, 0, size); | |
171 | ||
172 | return addr; | |
173 | } | |
174 | ||
175 | /* | |
176 | * highs, and targets are managed as dynamic arrays during a | |
177 | * table load. | |
178 | */ | |
179 | static int alloc_targets(struct dm_table *t, unsigned int num) | |
180 | { | |
181 | sector_t *n_highs; | |
182 | struct dm_target *n_targets; | |
183 | int n = t->num_targets; | |
184 | ||
185 | /* | |
186 | * Allocate both the target array and offset array at once. | |
187 | */ | |
188 | n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) + | |
189 | sizeof(sector_t)); | |
190 | if (!n_highs) | |
191 | return -ENOMEM; | |
192 | ||
193 | n_targets = (struct dm_target *) (n_highs + num); | |
194 | ||
195 | if (n) { | |
196 | memcpy(n_highs, t->highs, sizeof(*n_highs) * n); | |
197 | memcpy(n_targets, t->targets, sizeof(*n_targets) * n); | |
198 | } | |
199 | ||
200 | memset(n_highs + n, -1, sizeof(*n_highs) * (num - n)); | |
201 | vfree(t->highs); | |
202 | ||
203 | t->num_allocated = num; | |
204 | t->highs = n_highs; | |
205 | t->targets = n_targets; | |
206 | ||
207 | return 0; | |
208 | } | |
209 | ||
1134e5ae MA |
210 | int dm_table_create(struct dm_table **result, int mode, |
211 | unsigned num_targets, struct mapped_device *md) | |
1da177e4 LT |
212 | { |
213 | struct dm_table *t = kmalloc(sizeof(*t), GFP_KERNEL); | |
214 | ||
215 | if (!t) | |
216 | return -ENOMEM; | |
217 | ||
218 | memset(t, 0, sizeof(*t)); | |
219 | INIT_LIST_HEAD(&t->devices); | |
220 | atomic_set(&t->holders, 1); | |
221 | ||
222 | if (!num_targets) | |
223 | num_targets = KEYS_PER_NODE; | |
224 | ||
225 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | |
226 | ||
227 | if (alloc_targets(t, num_targets)) { | |
228 | kfree(t); | |
229 | t = NULL; | |
230 | return -ENOMEM; | |
231 | } | |
232 | ||
233 | t->mode = mode; | |
1134e5ae | 234 | t->md = md; |
1da177e4 LT |
235 | *result = t; |
236 | return 0; | |
237 | } | |
238 | ||
239 | static void free_devices(struct list_head *devices) | |
240 | { | |
241 | struct list_head *tmp, *next; | |
242 | ||
243 | for (tmp = devices->next; tmp != devices; tmp = next) { | |
244 | struct dm_dev *dd = list_entry(tmp, struct dm_dev, list); | |
245 | next = tmp->next; | |
246 | kfree(dd); | |
247 | } | |
248 | } | |
249 | ||
5e198d94 | 250 | static void table_destroy(struct dm_table *t) |
1da177e4 LT |
251 | { |
252 | unsigned int i; | |
253 | ||
254 | /* free the indexes (see dm_table_complete) */ | |
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 */ | |
271 | if (t->devices.next != &t->devices) { | |
272 | DMWARN("devices still present during destroy: " | |
273 | "dm_table_remove_device calls missing"); | |
274 | ||
275 | free_devices(&t->devices); | |
276 | } | |
277 | ||
278 | kfree(t); | |
279 | } | |
280 | ||
281 | void dm_table_get(struct dm_table *t) | |
282 | { | |
283 | atomic_inc(&t->holders); | |
284 | } | |
285 | ||
286 | void dm_table_put(struct dm_table *t) | |
287 | { | |
288 | if (!t) | |
289 | return; | |
290 | ||
291 | if (atomic_dec_and_test(&t->holders)) | |
292 | table_destroy(t); | |
293 | } | |
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 | ||
306 | /* | |
307 | * Convert a device path to a dev_t. | |
308 | */ | |
309 | static int lookup_device(const char *path, dev_t *dev) | |
310 | { | |
311 | int r; | |
312 | struct nameidata nd; | |
313 | struct inode *inode; | |
314 | ||
315 | if ((r = path_lookup(path, LOOKUP_FOLLOW, &nd))) | |
316 | return r; | |
317 | ||
318 | inode = nd.dentry->d_inode; | |
319 | if (!inode) { | |
320 | r = -ENOENT; | |
321 | goto out; | |
322 | } | |
323 | ||
324 | if (!S_ISBLK(inode->i_mode)) { | |
325 | r = -ENOTBLK; | |
326 | goto out; | |
327 | } | |
328 | ||
329 | *dev = inode->i_rdev; | |
330 | ||
331 | out: | |
332 | path_release(&nd); | |
333 | return r; | |
334 | } | |
335 | ||
336 | /* | |
337 | * See if we've already got a device in the list. | |
338 | */ | |
339 | static struct dm_dev *find_device(struct list_head *l, dev_t dev) | |
340 | { | |
341 | struct dm_dev *dd; | |
342 | ||
343 | list_for_each_entry (dd, l, list) | |
344 | if (dd->bdev->bd_dev == dev) | |
345 | return dd; | |
346 | ||
347 | return NULL; | |
348 | } | |
349 | ||
350 | /* | |
351 | * Open a device so we can use it as a map destination. | |
352 | */ | |
353 | static int open_dev(struct dm_dev *d, dev_t dev) | |
354 | { | |
355 | static char *_claim_ptr = "I belong to device-mapper"; | |
356 | struct block_device *bdev; | |
357 | ||
358 | int r; | |
359 | ||
547bc926 | 360 | BUG_ON(d->bdev); |
1da177e4 LT |
361 | |
362 | bdev = open_by_devnum(dev, d->mode); | |
363 | if (IS_ERR(bdev)) | |
364 | return PTR_ERR(bdev); | |
365 | r = bd_claim(bdev, _claim_ptr); | |
366 | if (r) | |
367 | blkdev_put(bdev); | |
368 | else | |
369 | d->bdev = bdev; | |
370 | return r; | |
371 | } | |
372 | ||
373 | /* | |
374 | * Close a device that we've been using. | |
375 | */ | |
376 | static void close_dev(struct dm_dev *d) | |
377 | { | |
378 | if (!d->bdev) | |
379 | return; | |
380 | ||
381 | bd_release(d->bdev); | |
382 | blkdev_put(d->bdev); | |
383 | d->bdev = NULL; | |
384 | } | |
385 | ||
386 | /* | |
387 | * If possible (ie. blk_size[major] is set), this checks an area | |
388 | * of a destination device is valid. | |
389 | */ | |
390 | static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len) | |
391 | { | |
392 | sector_t dev_size; | |
393 | dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT; | |
394 | return ((start < dev_size) && (len <= (dev_size - start))); | |
395 | } | |
396 | ||
397 | /* | |
398 | * This upgrades the mode on an already open dm_dev. Being | |
399 | * careful to leave things as they were if we fail to reopen the | |
400 | * device. | |
401 | */ | |
402 | static int upgrade_mode(struct dm_dev *dd, int new_mode) | |
403 | { | |
404 | int r; | |
405 | struct dm_dev dd_copy; | |
406 | dev_t dev = dd->bdev->bd_dev; | |
407 | ||
408 | dd_copy = *dd; | |
409 | ||
410 | dd->mode |= new_mode; | |
411 | dd->bdev = NULL; | |
412 | r = open_dev(dd, dev); | |
413 | if (!r) | |
414 | close_dev(&dd_copy); | |
415 | else | |
416 | *dd = dd_copy; | |
417 | ||
418 | return r; | |
419 | } | |
420 | ||
421 | /* | |
422 | * Add a device to the list, or just increment the usage count if | |
423 | * it's already present. | |
424 | */ | |
425 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | |
426 | const char *path, sector_t start, sector_t len, | |
427 | int mode, struct dm_dev **result) | |
428 | { | |
429 | int r; | |
430 | dev_t dev; | |
431 | struct dm_dev *dd; | |
432 | unsigned int major, minor; | |
433 | ||
547bc926 | 434 | BUG_ON(!t); |
1da177e4 LT |
435 | |
436 | if (sscanf(path, "%u:%u", &major, &minor) == 2) { | |
437 | /* Extract the major/minor numbers */ | |
438 | dev = MKDEV(major, minor); | |
439 | if (MAJOR(dev) != major || MINOR(dev) != minor) | |
440 | return -EOVERFLOW; | |
441 | } else { | |
442 | /* convert the path to a device */ | |
443 | if ((r = lookup_device(path, &dev))) | |
444 | return r; | |
445 | } | |
446 | ||
447 | dd = find_device(&t->devices, dev); | |
448 | if (!dd) { | |
449 | dd = kmalloc(sizeof(*dd), GFP_KERNEL); | |
450 | if (!dd) | |
451 | return -ENOMEM; | |
452 | ||
453 | dd->mode = mode; | |
454 | dd->bdev = NULL; | |
455 | ||
456 | if ((r = open_dev(dd, dev))) { | |
457 | kfree(dd); | |
458 | return r; | |
459 | } | |
460 | ||
461 | format_dev_t(dd->name, dev); | |
462 | ||
463 | atomic_set(&dd->count, 0); | |
464 | list_add(&dd->list, &t->devices); | |
465 | ||
466 | } else if (dd->mode != (mode | dd->mode)) { | |
467 | r = upgrade_mode(dd, mode); | |
468 | if (r) | |
469 | return r; | |
470 | } | |
471 | atomic_inc(&dd->count); | |
472 | ||
473 | if (!check_device_area(dd, start, len)) { | |
474 | DMWARN("device %s too small for target", path); | |
475 | dm_put_device(ti, dd); | |
476 | return -EINVAL; | |
477 | } | |
478 | ||
479 | *result = dd; | |
480 | ||
481 | return 0; | |
482 | } | |
483 | ||
484 | ||
485 | int dm_get_device(struct dm_target *ti, const char *path, sector_t start, | |
486 | sector_t len, int mode, struct dm_dev **result) | |
487 | { | |
488 | int r = __table_get_device(ti->table, ti, path, | |
489 | start, len, mode, result); | |
490 | if (!r) { | |
491 | request_queue_t *q = bdev_get_queue((*result)->bdev); | |
492 | struct io_restrictions *rs = &ti->limits; | |
493 | ||
494 | /* | |
495 | * Combine the device limits low. | |
496 | * | |
497 | * FIXME: if we move an io_restriction struct | |
498 | * into q this would just be a call to | |
499 | * combine_restrictions_low() | |
500 | */ | |
501 | rs->max_sectors = | |
502 | min_not_zero(rs->max_sectors, q->max_sectors); | |
503 | ||
504 | /* FIXME: Device-Mapper on top of RAID-0 breaks because DM | |
505 | * currently doesn't honor MD's merge_bvec_fn routine. | |
506 | * In this case, we'll force DM to use PAGE_SIZE or | |
507 | * smaller I/O, just to be safe. A better fix is in the | |
508 | * works, but add this for the time being so it will at | |
509 | * least operate correctly. | |
510 | */ | |
511 | if (q->merge_bvec_fn) | |
512 | rs->max_sectors = | |
513 | min_not_zero(rs->max_sectors, | |
3ee247eb | 514 | (unsigned int) (PAGE_SIZE >> 9)); |
1da177e4 LT |
515 | |
516 | rs->max_phys_segments = | |
517 | min_not_zero(rs->max_phys_segments, | |
518 | q->max_phys_segments); | |
519 | ||
520 | rs->max_hw_segments = | |
521 | min_not_zero(rs->max_hw_segments, q->max_hw_segments); | |
522 | ||
523 | rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size); | |
524 | ||
525 | rs->max_segment_size = | |
526 | min_not_zero(rs->max_segment_size, q->max_segment_size); | |
527 | ||
528 | rs->seg_boundary_mask = | |
529 | min_not_zero(rs->seg_boundary_mask, | |
530 | q->seg_boundary_mask); | |
969429b5 N |
531 | |
532 | rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags); | |
1da177e4 LT |
533 | } |
534 | ||
535 | return r; | |
536 | } | |
537 | ||
538 | /* | |
539 | * Decrement a devices use count and remove it if necessary. | |
540 | */ | |
541 | void dm_put_device(struct dm_target *ti, struct dm_dev *dd) | |
542 | { | |
543 | if (atomic_dec_and_test(&dd->count)) { | |
544 | close_dev(dd); | |
545 | list_del(&dd->list); | |
546 | kfree(dd); | |
547 | } | |
548 | } | |
549 | ||
550 | /* | |
551 | * Checks to see if the target joins onto the end of the table. | |
552 | */ | |
553 | static int adjoin(struct dm_table *table, struct dm_target *ti) | |
554 | { | |
555 | struct dm_target *prev; | |
556 | ||
557 | if (!table->num_targets) | |
558 | return !ti->begin; | |
559 | ||
560 | prev = &table->targets[table->num_targets - 1]; | |
561 | return (ti->begin == (prev->begin + prev->len)); | |
562 | } | |
563 | ||
564 | /* | |
565 | * Used to dynamically allocate the arg array. | |
566 | */ | |
567 | static char **realloc_argv(unsigned *array_size, char **old_argv) | |
568 | { | |
569 | char **argv; | |
570 | unsigned new_size; | |
571 | ||
572 | new_size = *array_size ? *array_size * 2 : 64; | |
573 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | |
574 | if (argv) { | |
575 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | |
576 | *array_size = new_size; | |
577 | } | |
578 | ||
579 | kfree(old_argv); | |
580 | return argv; | |
581 | } | |
582 | ||
583 | /* | |
584 | * Destructively splits up the argument list to pass to ctr. | |
585 | */ | |
586 | int dm_split_args(int *argc, char ***argvp, char *input) | |
587 | { | |
588 | char *start, *end = input, *out, **argv = NULL; | |
589 | unsigned array_size = 0; | |
590 | ||
591 | *argc = 0; | |
592 | argv = realloc_argv(&array_size, argv); | |
593 | if (!argv) | |
594 | return -ENOMEM; | |
595 | ||
596 | while (1) { | |
597 | start = end; | |
598 | ||
599 | /* Skip whitespace */ | |
600 | while (*start && isspace(*start)) | |
601 | start++; | |
602 | ||
603 | if (!*start) | |
604 | break; /* success, we hit the end */ | |
605 | ||
606 | /* 'out' is used to remove any back-quotes */ | |
607 | end = out = start; | |
608 | while (*end) { | |
609 | /* Everything apart from '\0' can be quoted */ | |
610 | if (*end == '\\' && *(end + 1)) { | |
611 | *out++ = *(end + 1); | |
612 | end += 2; | |
613 | continue; | |
614 | } | |
615 | ||
616 | if (isspace(*end)) | |
617 | break; /* end of token */ | |
618 | ||
619 | *out++ = *end++; | |
620 | } | |
621 | ||
622 | /* have we already filled the array ? */ | |
623 | if ((*argc + 1) > array_size) { | |
624 | argv = realloc_argv(&array_size, argv); | |
625 | if (!argv) | |
626 | return -ENOMEM; | |
627 | } | |
628 | ||
629 | /* we know this is whitespace */ | |
630 | if (*end) | |
631 | end++; | |
632 | ||
633 | /* terminate the string and put it in the array */ | |
634 | *out = '\0'; | |
635 | argv[*argc] = start; | |
636 | (*argc)++; | |
637 | } | |
638 | ||
639 | *argvp = argv; | |
640 | return 0; | |
641 | } | |
642 | ||
643 | static void check_for_valid_limits(struct io_restrictions *rs) | |
644 | { | |
645 | if (!rs->max_sectors) | |
defd94b7 | 646 | rs->max_sectors = SAFE_MAX_SECTORS; |
1da177e4 LT |
647 | if (!rs->max_phys_segments) |
648 | rs->max_phys_segments = MAX_PHYS_SEGMENTS; | |
649 | if (!rs->max_hw_segments) | |
650 | rs->max_hw_segments = MAX_HW_SEGMENTS; | |
651 | if (!rs->hardsect_size) | |
652 | rs->hardsect_size = 1 << SECTOR_SHIFT; | |
653 | if (!rs->max_segment_size) | |
654 | rs->max_segment_size = MAX_SEGMENT_SIZE; | |
655 | if (!rs->seg_boundary_mask) | |
656 | rs->seg_boundary_mask = -1; | |
657 | } | |
658 | ||
659 | int dm_table_add_target(struct dm_table *t, const char *type, | |
660 | sector_t start, sector_t len, char *params) | |
661 | { | |
662 | int r = -EINVAL, argc; | |
663 | char **argv; | |
664 | struct dm_target *tgt; | |
665 | ||
666 | if ((r = check_space(t))) | |
667 | return r; | |
668 | ||
669 | tgt = t->targets + t->num_targets; | |
670 | memset(tgt, 0, sizeof(*tgt)); | |
671 | ||
672 | if (!len) { | |
673 | tgt->error = "zero-length target"; | |
674 | DMERR("%s", tgt->error); | |
675 | return -EINVAL; | |
676 | } | |
677 | ||
678 | tgt->type = dm_get_target_type(type); | |
679 | if (!tgt->type) { | |
680 | tgt->error = "unknown target type"; | |
681 | DMERR("%s", tgt->error); | |
682 | return -EINVAL; | |
683 | } | |
684 | ||
685 | tgt->table = t; | |
686 | tgt->begin = start; | |
687 | tgt->len = len; | |
688 | tgt->error = "Unknown error"; | |
689 | ||
690 | /* | |
691 | * Does this target adjoin the previous one ? | |
692 | */ | |
693 | if (!adjoin(t, tgt)) { | |
694 | tgt->error = "Gap in table"; | |
695 | r = -EINVAL; | |
696 | goto bad; | |
697 | } | |
698 | ||
699 | r = dm_split_args(&argc, &argv, params); | |
700 | if (r) { | |
701 | tgt->error = "couldn't split parameters (insufficient memory)"; | |
702 | goto bad; | |
703 | } | |
704 | ||
705 | r = tgt->type->ctr(tgt, argc, argv); | |
706 | kfree(argv); | |
707 | if (r) | |
708 | goto bad; | |
709 | ||
710 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | |
711 | ||
712 | /* FIXME: the plan is to combine high here and then have | |
713 | * the merge fn apply the target level restrictions. */ | |
714 | combine_restrictions_low(&t->limits, &tgt->limits); | |
715 | return 0; | |
716 | ||
717 | bad: | |
718 | DMERR("%s", tgt->error); | |
719 | dm_put_target_type(tgt->type); | |
720 | return r; | |
721 | } | |
722 | ||
723 | static int setup_indexes(struct dm_table *t) | |
724 | { | |
725 | int i; | |
726 | unsigned int total = 0; | |
727 | sector_t *indexes; | |
728 | ||
729 | /* allocate the space for *all* the indexes */ | |
730 | for (i = t->depth - 2; i >= 0; i--) { | |
731 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | |
732 | total += t->counts[i]; | |
733 | } | |
734 | ||
735 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | |
736 | if (!indexes) | |
737 | return -ENOMEM; | |
738 | ||
739 | /* set up internal nodes, bottom-up */ | |
740 | for (i = t->depth - 2, total = 0; i >= 0; i--) { | |
741 | t->index[i] = indexes; | |
742 | indexes += (KEYS_PER_NODE * t->counts[i]); | |
743 | setup_btree_index(i, t); | |
744 | } | |
745 | ||
746 | return 0; | |
747 | } | |
748 | ||
749 | /* | |
750 | * Builds the btree to index the map. | |
751 | */ | |
752 | int dm_table_complete(struct dm_table *t) | |
753 | { | |
754 | int r = 0; | |
755 | unsigned int leaf_nodes; | |
756 | ||
757 | check_for_valid_limits(&t->limits); | |
758 | ||
759 | /* how many indexes will the btree have ? */ | |
760 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | |
761 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | |
762 | ||
763 | /* leaf layer has already been set up */ | |
764 | t->counts[t->depth - 1] = leaf_nodes; | |
765 | t->index[t->depth - 1] = t->highs; | |
766 | ||
767 | if (t->depth >= 2) | |
768 | r = setup_indexes(t); | |
769 | ||
770 | return r; | |
771 | } | |
772 | ||
773 | static DECLARE_MUTEX(_event_lock); | |
774 | void dm_table_event_callback(struct dm_table *t, | |
775 | void (*fn)(void *), void *context) | |
776 | { | |
777 | down(&_event_lock); | |
778 | t->event_fn = fn; | |
779 | t->event_context = context; | |
780 | up(&_event_lock); | |
781 | } | |
782 | ||
783 | void dm_table_event(struct dm_table *t) | |
784 | { | |
785 | /* | |
786 | * You can no longer call dm_table_event() from interrupt | |
787 | * context, use a bottom half instead. | |
788 | */ | |
789 | BUG_ON(in_interrupt()); | |
790 | ||
791 | down(&_event_lock); | |
792 | if (t->event_fn) | |
793 | t->event_fn(t->event_context); | |
794 | up(&_event_lock); | |
795 | } | |
796 | ||
797 | sector_t dm_table_get_size(struct dm_table *t) | |
798 | { | |
799 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | |
800 | } | |
801 | ||
802 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | |
803 | { | |
804 | if (index > t->num_targets) | |
805 | return NULL; | |
806 | ||
807 | return t->targets + index; | |
808 | } | |
809 | ||
810 | /* | |
811 | * Search the btree for the correct target. | |
812 | */ | |
813 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | |
814 | { | |
815 | unsigned int l, n = 0, k = 0; | |
816 | sector_t *node; | |
817 | ||
818 | for (l = 0; l < t->depth; l++) { | |
819 | n = get_child(n, k); | |
820 | node = get_node(t, l, n); | |
821 | ||
822 | for (k = 0; k < KEYS_PER_NODE; k++) | |
823 | if (node[k] >= sector) | |
824 | break; | |
825 | } | |
826 | ||
827 | return &t->targets[(KEYS_PER_NODE * n) + k]; | |
828 | } | |
829 | ||
830 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q) | |
831 | { | |
832 | /* | |
833 | * Make sure we obey the optimistic sub devices | |
834 | * restrictions. | |
835 | */ | |
836 | blk_queue_max_sectors(q, t->limits.max_sectors); | |
837 | q->max_phys_segments = t->limits.max_phys_segments; | |
838 | q->max_hw_segments = t->limits.max_hw_segments; | |
839 | q->hardsect_size = t->limits.hardsect_size; | |
840 | q->max_segment_size = t->limits.max_segment_size; | |
841 | q->seg_boundary_mask = t->limits.seg_boundary_mask; | |
969429b5 N |
842 | if (t->limits.no_cluster) |
843 | q->queue_flags &= ~(1 << QUEUE_FLAG_CLUSTER); | |
844 | else | |
845 | q->queue_flags |= (1 << QUEUE_FLAG_CLUSTER); | |
846 | ||
1da177e4 LT |
847 | } |
848 | ||
849 | unsigned int dm_table_get_num_targets(struct dm_table *t) | |
850 | { | |
851 | return t->num_targets; | |
852 | } | |
853 | ||
854 | struct list_head *dm_table_get_devices(struct dm_table *t) | |
855 | { | |
856 | return &t->devices; | |
857 | } | |
858 | ||
859 | int dm_table_get_mode(struct dm_table *t) | |
860 | { | |
861 | return t->mode; | |
862 | } | |
863 | ||
864 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | |
865 | { | |
866 | int i = t->num_targets; | |
867 | struct dm_target *ti = t->targets; | |
868 | ||
869 | while (i--) { | |
870 | if (postsuspend) { | |
871 | if (ti->type->postsuspend) | |
872 | ti->type->postsuspend(ti); | |
873 | } else if (ti->type->presuspend) | |
874 | ti->type->presuspend(ti); | |
875 | ||
876 | ti++; | |
877 | } | |
878 | } | |
879 | ||
880 | void dm_table_presuspend_targets(struct dm_table *t) | |
881 | { | |
cf222b37 AK |
882 | if (!t) |
883 | return; | |
884 | ||
1da177e4 LT |
885 | return suspend_targets(t, 0); |
886 | } | |
887 | ||
888 | void dm_table_postsuspend_targets(struct dm_table *t) | |
889 | { | |
cf222b37 AK |
890 | if (!t) |
891 | return; | |
892 | ||
1da177e4 LT |
893 | return suspend_targets(t, 1); |
894 | } | |
895 | ||
896 | void dm_table_resume_targets(struct dm_table *t) | |
897 | { | |
898 | int i; | |
899 | ||
900 | for (i = 0; i < t->num_targets; i++) { | |
901 | struct dm_target *ti = t->targets + i; | |
902 | ||
903 | if (ti->type->resume) | |
904 | ti->type->resume(ti); | |
905 | } | |
906 | } | |
907 | ||
908 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | |
909 | { | |
910 | struct list_head *d, *devices; | |
911 | int r = 0; | |
912 | ||
913 | devices = dm_table_get_devices(t); | |
914 | for (d = devices->next; d != devices; d = d->next) { | |
915 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | |
916 | request_queue_t *q = bdev_get_queue(dd->bdev); | |
917 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | |
918 | } | |
919 | ||
920 | return r; | |
921 | } | |
922 | ||
923 | void dm_table_unplug_all(struct dm_table *t) | |
924 | { | |
925 | struct list_head *d, *devices = dm_table_get_devices(t); | |
926 | ||
927 | for (d = devices->next; d != devices; d = d->next) { | |
928 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | |
929 | request_queue_t *q = bdev_get_queue(dd->bdev); | |
930 | ||
931 | if (q->unplug_fn) | |
932 | q->unplug_fn(q); | |
933 | } | |
934 | } | |
935 | ||
936 | int dm_table_flush_all(struct dm_table *t) | |
937 | { | |
938 | struct list_head *d, *devices = dm_table_get_devices(t); | |
939 | int ret = 0; | |
940 | ||
941 | for (d = devices->next; d != devices; d = d->next) { | |
942 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | |
943 | request_queue_t *q = bdev_get_queue(dd->bdev); | |
944 | int err; | |
945 | ||
946 | if (!q->issue_flush_fn) | |
947 | err = -EOPNOTSUPP; | |
948 | else | |
949 | err = q->issue_flush_fn(q, dd->bdev->bd_disk, NULL); | |
950 | ||
951 | if (!ret) | |
952 | ret = err; | |
953 | } | |
954 | ||
955 | return ret; | |
956 | } | |
957 | ||
1134e5ae MA |
958 | struct mapped_device *dm_table_get_md(struct dm_table *t) |
959 | { | |
960 | dm_get(t->md); | |
961 | ||
962 | return t->md; | |
963 | } | |
964 | ||
1da177e4 LT |
965 | EXPORT_SYMBOL(dm_vcalloc); |
966 | EXPORT_SYMBOL(dm_get_device); | |
967 | EXPORT_SYMBOL(dm_put_device); | |
968 | EXPORT_SYMBOL(dm_table_event); | |
d5e404c1 | 969 | EXPORT_SYMBOL(dm_table_get_size); |
1da177e4 | 970 | EXPORT_SYMBOL(dm_table_get_mode); |
1134e5ae | 971 | EXPORT_SYMBOL(dm_table_get_md); |
1da177e4 LT |
972 | EXPORT_SYMBOL(dm_table_put); |
973 | EXPORT_SYMBOL(dm_table_get); | |
974 | EXPORT_SYMBOL(dm_table_unplug_all); | |
975 | EXPORT_SYMBOL(dm_table_flush_all); |