diff options
author | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 18:20:36 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 18:20:36 -0400 |
commit | 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch) | |
tree | 0bba044c4ce775e45a88a51686b5d9f90697ea9d /drivers/md/dm-table.c |
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Diffstat (limited to 'drivers/md/dm-table.c')
-rw-r--r-- | drivers/md/dm-table.c | 950 |
1 files changed, 950 insertions, 0 deletions
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c new file mode 100644 index 000000000000..ee175d4906c4 --- /dev/null +++ b/drivers/md/dm-table.c | |||
@@ -0,0 +1,950 @@ | |||
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 { | ||
25 | atomic_t holders; | ||
26 | |||
27 | /* btree table */ | ||
28 | unsigned int depth; | ||
29 | unsigned int counts[MAX_DEPTH]; /* in nodes */ | ||
30 | sector_t *index[MAX_DEPTH]; | ||
31 | |||
32 | unsigned int num_targets; | ||
33 | unsigned int num_allocated; | ||
34 | sector_t *highs; | ||
35 | struct dm_target *targets; | ||
36 | |||
37 | /* | ||
38 | * Indicates the rw permissions for the new logical | ||
39 | * device. This should be a combination of FMODE_READ | ||
40 | * and FMODE_WRITE. | ||
41 | */ | ||
42 | int mode; | ||
43 | |||
44 | /* a list of devices used by this table */ | ||
45 | struct list_head devices; | ||
46 | |||
47 | /* | ||
48 | * These are optimistic limits taken from all the | ||
49 | * targets, some targets will need smaller limits. | ||
50 | */ | ||
51 | struct io_restrictions limits; | ||
52 | |||
53 | /* events get handed up using this callback */ | ||
54 | void (*event_fn)(void *); | ||
55 | void *event_context; | ||
56 | }; | ||
57 | |||
58 | /* | ||
59 | * Similar to ceiling(log_size(n)) | ||
60 | */ | ||
61 | static unsigned int int_log(unsigned int n, unsigned int base) | ||
62 | { | ||
63 | int result = 0; | ||
64 | |||
65 | while (n > 1) { | ||
66 | n = dm_div_up(n, base); | ||
67 | result++; | ||
68 | } | ||
69 | |||
70 | return result; | ||
71 | } | ||
72 | |||
73 | /* | ||
74 | * Returns the minimum that is _not_ zero, unless both are zero. | ||
75 | */ | ||
76 | #define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r)) | ||
77 | |||
78 | /* | ||
79 | * Combine two io_restrictions, always taking the lower value. | ||
80 | */ | ||
81 | static void combine_restrictions_low(struct io_restrictions *lhs, | ||
82 | struct io_restrictions *rhs) | ||
83 | { | ||
84 | lhs->max_sectors = | ||
85 | min_not_zero(lhs->max_sectors, rhs->max_sectors); | ||
86 | |||
87 | lhs->max_phys_segments = | ||
88 | min_not_zero(lhs->max_phys_segments, rhs->max_phys_segments); | ||
89 | |||
90 | lhs->max_hw_segments = | ||
91 | min_not_zero(lhs->max_hw_segments, rhs->max_hw_segments); | ||
92 | |||
93 | lhs->hardsect_size = max(lhs->hardsect_size, rhs->hardsect_size); | ||
94 | |||
95 | lhs->max_segment_size = | ||
96 | min_not_zero(lhs->max_segment_size, rhs->max_segment_size); | ||
97 | |||
98 | lhs->seg_boundary_mask = | ||
99 | min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask); | ||
100 | } | ||
101 | |||
102 | /* | ||
103 | * Calculate the index of the child node of the n'th node k'th key. | ||
104 | */ | ||
105 | static inline unsigned int get_child(unsigned int n, unsigned int k) | ||
106 | { | ||
107 | return (n * CHILDREN_PER_NODE) + k; | ||
108 | } | ||
109 | |||
110 | /* | ||
111 | * Return the n'th node of level l from table t. | ||
112 | */ | ||
113 | static inline sector_t *get_node(struct dm_table *t, | ||
114 | unsigned int l, unsigned int n) | ||
115 | { | ||
116 | return t->index[l] + (n * KEYS_PER_NODE); | ||
117 | } | ||
118 | |||
119 | /* | ||
120 | * Return the highest key that you could lookup from the n'th | ||
121 | * node on level l of the btree. | ||
122 | */ | ||
123 | static sector_t high(struct dm_table *t, unsigned int l, unsigned int n) | ||
124 | { | ||
125 | for (; l < t->depth - 1; l++) | ||
126 | n = get_child(n, CHILDREN_PER_NODE - 1); | ||
127 | |||
128 | if (n >= t->counts[l]) | ||
129 | return (sector_t) - 1; | ||
130 | |||
131 | return get_node(t, l, n)[KEYS_PER_NODE - 1]; | ||
132 | } | ||
133 | |||
134 | /* | ||
135 | * Fills in a level of the btree based on the highs of the level | ||
136 | * below it. | ||
137 | */ | ||
138 | static int setup_btree_index(unsigned int l, struct dm_table *t) | ||
139 | { | ||
140 | unsigned int n, k; | ||
141 | sector_t *node; | ||
142 | |||
143 | for (n = 0U; n < t->counts[l]; n++) { | ||
144 | node = get_node(t, l, n); | ||
145 | |||
146 | for (k = 0U; k < KEYS_PER_NODE; k++) | ||
147 | node[k] = high(t, l + 1, get_child(n, k)); | ||
148 | } | ||
149 | |||
150 | return 0; | ||
151 | } | ||
152 | |||
153 | void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size) | ||
154 | { | ||
155 | unsigned long size; | ||
156 | void *addr; | ||
157 | |||
158 | /* | ||
159 | * Check that we're not going to overflow. | ||
160 | */ | ||
161 | if (nmemb > (ULONG_MAX / elem_size)) | ||
162 | return NULL; | ||
163 | |||
164 | size = nmemb * elem_size; | ||
165 | addr = vmalloc(size); | ||
166 | if (addr) | ||
167 | memset(addr, 0, size); | ||
168 | |||
169 | return addr; | ||
170 | } | ||
171 | |||
172 | /* | ||
173 | * highs, and targets are managed as dynamic arrays during a | ||
174 | * table load. | ||
175 | */ | ||
176 | static int alloc_targets(struct dm_table *t, unsigned int num) | ||
177 | { | ||
178 | sector_t *n_highs; | ||
179 | struct dm_target *n_targets; | ||
180 | int n = t->num_targets; | ||
181 | |||
182 | /* | ||
183 | * Allocate both the target array and offset array at once. | ||
184 | */ | ||
185 | n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) + | ||
186 | sizeof(sector_t)); | ||
187 | if (!n_highs) | ||
188 | return -ENOMEM; | ||
189 | |||
190 | n_targets = (struct dm_target *) (n_highs + num); | ||
191 | |||
192 | if (n) { | ||
193 | memcpy(n_highs, t->highs, sizeof(*n_highs) * n); | ||
194 | memcpy(n_targets, t->targets, sizeof(*n_targets) * n); | ||
195 | } | ||
196 | |||
197 | memset(n_highs + n, -1, sizeof(*n_highs) * (num - n)); | ||
198 | vfree(t->highs); | ||
199 | |||
200 | t->num_allocated = num; | ||
201 | t->highs = n_highs; | ||
202 | t->targets = n_targets; | ||
203 | |||
204 | return 0; | ||
205 | } | ||
206 | |||
207 | int dm_table_create(struct dm_table **result, int mode, unsigned num_targets) | ||
208 | { | ||
209 | struct dm_table *t = kmalloc(sizeof(*t), GFP_KERNEL); | ||
210 | |||
211 | if (!t) | ||
212 | return -ENOMEM; | ||
213 | |||
214 | memset(t, 0, sizeof(*t)); | ||
215 | INIT_LIST_HEAD(&t->devices); | ||
216 | atomic_set(&t->holders, 1); | ||
217 | |||
218 | if (!num_targets) | ||
219 | num_targets = KEYS_PER_NODE; | ||
220 | |||
221 | num_targets = dm_round_up(num_targets, KEYS_PER_NODE); | ||
222 | |||
223 | if (alloc_targets(t, num_targets)) { | ||
224 | kfree(t); | ||
225 | t = NULL; | ||
226 | return -ENOMEM; | ||
227 | } | ||
228 | |||
229 | t->mode = mode; | ||
230 | *result = t; | ||
231 | return 0; | ||
232 | } | ||
233 | |||
234 | static void free_devices(struct list_head *devices) | ||
235 | { | ||
236 | struct list_head *tmp, *next; | ||
237 | |||
238 | for (tmp = devices->next; tmp != devices; tmp = next) { | ||
239 | struct dm_dev *dd = list_entry(tmp, struct dm_dev, list); | ||
240 | next = tmp->next; | ||
241 | kfree(dd); | ||
242 | } | ||
243 | } | ||
244 | |||
245 | void table_destroy(struct dm_table *t) | ||
246 | { | ||
247 | unsigned int i; | ||
248 | |||
249 | /* free the indexes (see dm_table_complete) */ | ||
250 | if (t->depth >= 2) | ||
251 | vfree(t->index[t->depth - 2]); | ||
252 | |||
253 | /* free the targets */ | ||
254 | for (i = 0; i < t->num_targets; i++) { | ||
255 | struct dm_target *tgt = t->targets + i; | ||
256 | |||
257 | if (tgt->type->dtr) | ||
258 | tgt->type->dtr(tgt); | ||
259 | |||
260 | dm_put_target_type(tgt->type); | ||
261 | } | ||
262 | |||
263 | vfree(t->highs); | ||
264 | |||
265 | /* free the device list */ | ||
266 | if (t->devices.next != &t->devices) { | ||
267 | DMWARN("devices still present during destroy: " | ||
268 | "dm_table_remove_device calls missing"); | ||
269 | |||
270 | free_devices(&t->devices); | ||
271 | } | ||
272 | |||
273 | kfree(t); | ||
274 | } | ||
275 | |||
276 | void dm_table_get(struct dm_table *t) | ||
277 | { | ||
278 | atomic_inc(&t->holders); | ||
279 | } | ||
280 | |||
281 | void dm_table_put(struct dm_table *t) | ||
282 | { | ||
283 | if (!t) | ||
284 | return; | ||
285 | |||
286 | if (atomic_dec_and_test(&t->holders)) | ||
287 | table_destroy(t); | ||
288 | } | ||
289 | |||
290 | /* | ||
291 | * Checks to see if we need to extend highs or targets. | ||
292 | */ | ||
293 | static inline int check_space(struct dm_table *t) | ||
294 | { | ||
295 | if (t->num_targets >= t->num_allocated) | ||
296 | return alloc_targets(t, t->num_allocated * 2); | ||
297 | |||
298 | return 0; | ||
299 | } | ||
300 | |||
301 | /* | ||
302 | * Convert a device path to a dev_t. | ||
303 | */ | ||
304 | static int lookup_device(const char *path, dev_t *dev) | ||
305 | { | ||
306 | int r; | ||
307 | struct nameidata nd; | ||
308 | struct inode *inode; | ||
309 | |||
310 | if ((r = path_lookup(path, LOOKUP_FOLLOW, &nd))) | ||
311 | return r; | ||
312 | |||
313 | inode = nd.dentry->d_inode; | ||
314 | if (!inode) { | ||
315 | r = -ENOENT; | ||
316 | goto out; | ||
317 | } | ||
318 | |||
319 | if (!S_ISBLK(inode->i_mode)) { | ||
320 | r = -ENOTBLK; | ||
321 | goto out; | ||
322 | } | ||
323 | |||
324 | *dev = inode->i_rdev; | ||
325 | |||
326 | out: | ||
327 | path_release(&nd); | ||
328 | return r; | ||
329 | } | ||
330 | |||
331 | /* | ||
332 | * See if we've already got a device in the list. | ||
333 | */ | ||
334 | static struct dm_dev *find_device(struct list_head *l, dev_t dev) | ||
335 | { | ||
336 | struct dm_dev *dd; | ||
337 | |||
338 | list_for_each_entry (dd, l, list) | ||
339 | if (dd->bdev->bd_dev == dev) | ||
340 | return dd; | ||
341 | |||
342 | return NULL; | ||
343 | } | ||
344 | |||
345 | /* | ||
346 | * Open a device so we can use it as a map destination. | ||
347 | */ | ||
348 | static int open_dev(struct dm_dev *d, dev_t dev) | ||
349 | { | ||
350 | static char *_claim_ptr = "I belong to device-mapper"; | ||
351 | struct block_device *bdev; | ||
352 | |||
353 | int r; | ||
354 | |||
355 | if (d->bdev) | ||
356 | BUG(); | ||
357 | |||
358 | bdev = open_by_devnum(dev, d->mode); | ||
359 | if (IS_ERR(bdev)) | ||
360 | return PTR_ERR(bdev); | ||
361 | r = bd_claim(bdev, _claim_ptr); | ||
362 | if (r) | ||
363 | blkdev_put(bdev); | ||
364 | else | ||
365 | d->bdev = bdev; | ||
366 | return r; | ||
367 | } | ||
368 | |||
369 | /* | ||
370 | * Close a device that we've been using. | ||
371 | */ | ||
372 | static void close_dev(struct dm_dev *d) | ||
373 | { | ||
374 | if (!d->bdev) | ||
375 | return; | ||
376 | |||
377 | bd_release(d->bdev); | ||
378 | blkdev_put(d->bdev); | ||
379 | d->bdev = NULL; | ||
380 | } | ||
381 | |||
382 | /* | ||
383 | * If possible (ie. blk_size[major] is set), this checks an area | ||
384 | * of a destination device is valid. | ||
385 | */ | ||
386 | static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len) | ||
387 | { | ||
388 | sector_t dev_size; | ||
389 | dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT; | ||
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 | */ | ||
398 | static int upgrade_mode(struct dm_dev *dd, int new_mode) | ||
399 | { | ||
400 | int r; | ||
401 | struct dm_dev dd_copy; | ||
402 | dev_t dev = dd->bdev->bd_dev; | ||
403 | |||
404 | dd_copy = *dd; | ||
405 | |||
406 | dd->mode |= new_mode; | ||
407 | dd->bdev = NULL; | ||
408 | r = open_dev(dd, dev); | ||
409 | if (!r) | ||
410 | close_dev(&dd_copy); | ||
411 | else | ||
412 | *dd = dd_copy; | ||
413 | |||
414 | return r; | ||
415 | } | ||
416 | |||
417 | /* | ||
418 | * Add a device to the list, or just increment the usage count if | ||
419 | * it's already present. | ||
420 | */ | ||
421 | static int __table_get_device(struct dm_table *t, struct dm_target *ti, | ||
422 | const char *path, sector_t start, sector_t len, | ||
423 | int mode, struct dm_dev **result) | ||
424 | { | ||
425 | int r; | ||
426 | dev_t dev; | ||
427 | struct dm_dev *dd; | ||
428 | unsigned int major, minor; | ||
429 | |||
430 | if (!t) | ||
431 | BUG(); | ||
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 | |||
450 | dd->mode = mode; | ||
451 | dd->bdev = NULL; | ||
452 | |||
453 | if ((r = open_dev(dd, dev))) { | ||
454 | kfree(dd); | ||
455 | return r; | ||
456 | } | ||
457 | |||
458 | format_dev_t(dd->name, dev); | ||
459 | |||
460 | atomic_set(&dd->count, 0); | ||
461 | list_add(&dd->list, &t->devices); | ||
462 | |||
463 | } else if (dd->mode != (mode | dd->mode)) { | ||
464 | r = upgrade_mode(dd, mode); | ||
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); | ||
472 | dm_put_device(ti, dd); | ||
473 | return -EINVAL; | ||
474 | } | ||
475 | |||
476 | *result = dd; | ||
477 | |||
478 | return 0; | ||
479 | } | ||
480 | |||
481 | |||
482 | int dm_get_device(struct dm_target *ti, const char *path, sector_t start, | ||
483 | sector_t len, int mode, struct dm_dev **result) | ||
484 | { | ||
485 | int r = __table_get_device(ti->table, ti, path, | ||
486 | start, len, mode, result); | ||
487 | if (!r) { | ||
488 | request_queue_t *q = bdev_get_queue((*result)->bdev); | ||
489 | struct io_restrictions *rs = &ti->limits; | ||
490 | |||
491 | /* | ||
492 | * Combine the device limits low. | ||
493 | * | ||
494 | * FIXME: if we move an io_restriction struct | ||
495 | * into q this would just be a call to | ||
496 | * combine_restrictions_low() | ||
497 | */ | ||
498 | rs->max_sectors = | ||
499 | min_not_zero(rs->max_sectors, q->max_sectors); | ||
500 | |||
501 | /* FIXME: Device-Mapper on top of RAID-0 breaks because DM | ||
502 | * currently doesn't honor MD's merge_bvec_fn routine. | ||
503 | * In this case, we'll force DM to use PAGE_SIZE or | ||
504 | * smaller I/O, just to be safe. A better fix is in the | ||
505 | * works, but add this for the time being so it will at | ||
506 | * least operate correctly. | ||
507 | */ | ||
508 | if (q->merge_bvec_fn) | ||
509 | rs->max_sectors = | ||
510 | min_not_zero(rs->max_sectors, | ||
511 | (unsigned short)(PAGE_SIZE >> 9)); | ||
512 | |||
513 | rs->max_phys_segments = | ||
514 | min_not_zero(rs->max_phys_segments, | ||
515 | q->max_phys_segments); | ||
516 | |||
517 | rs->max_hw_segments = | ||
518 | min_not_zero(rs->max_hw_segments, q->max_hw_segments); | ||
519 | |||
520 | rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size); | ||
521 | |||
522 | rs->max_segment_size = | ||
523 | min_not_zero(rs->max_segment_size, q->max_segment_size); | ||
524 | |||
525 | rs->seg_boundary_mask = | ||
526 | min_not_zero(rs->seg_boundary_mask, | ||
527 | q->seg_boundary_mask); | ||
528 | } | ||
529 | |||
530 | return r; | ||
531 | } | ||
532 | |||
533 | /* | ||
534 | * Decrement a devices use count and remove it if necessary. | ||
535 | */ | ||
536 | void dm_put_device(struct dm_target *ti, struct dm_dev *dd) | ||
537 | { | ||
538 | if (atomic_dec_and_test(&dd->count)) { | ||
539 | close_dev(dd); | ||
540 | list_del(&dd->list); | ||
541 | kfree(dd); | ||
542 | } | ||
543 | } | ||
544 | |||
545 | /* | ||
546 | * Checks to see if the target joins onto the end of the table. | ||
547 | */ | ||
548 | static int adjoin(struct dm_table *table, struct dm_target *ti) | ||
549 | { | ||
550 | struct dm_target *prev; | ||
551 | |||
552 | if (!table->num_targets) | ||
553 | return !ti->begin; | ||
554 | |||
555 | prev = &table->targets[table->num_targets - 1]; | ||
556 | return (ti->begin == (prev->begin + prev->len)); | ||
557 | } | ||
558 | |||
559 | /* | ||
560 | * Used to dynamically allocate the arg array. | ||
561 | */ | ||
562 | static char **realloc_argv(unsigned *array_size, char **old_argv) | ||
563 | { | ||
564 | char **argv; | ||
565 | unsigned new_size; | ||
566 | |||
567 | new_size = *array_size ? *array_size * 2 : 64; | ||
568 | argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL); | ||
569 | if (argv) { | ||
570 | memcpy(argv, old_argv, *array_size * sizeof(*argv)); | ||
571 | *array_size = new_size; | ||
572 | } | ||
573 | |||
574 | kfree(old_argv); | ||
575 | return argv; | ||
576 | } | ||
577 | |||
578 | /* | ||
579 | * Destructively splits up the argument list to pass to ctr. | ||
580 | */ | ||
581 | int dm_split_args(int *argc, char ***argvp, char *input) | ||
582 | { | ||
583 | char *start, *end = input, *out, **argv = NULL; | ||
584 | unsigned array_size = 0; | ||
585 | |||
586 | *argc = 0; | ||
587 | argv = realloc_argv(&array_size, argv); | ||
588 | if (!argv) | ||
589 | return -ENOMEM; | ||
590 | |||
591 | while (1) { | ||
592 | start = end; | ||
593 | |||
594 | /* Skip whitespace */ | ||
595 | while (*start && isspace(*start)) | ||
596 | start++; | ||
597 | |||
598 | if (!*start) | ||
599 | break; /* success, we hit the end */ | ||
600 | |||
601 | /* 'out' is used to remove any back-quotes */ | ||
602 | end = out = start; | ||
603 | while (*end) { | ||
604 | /* Everything apart from '\0' can be quoted */ | ||
605 | if (*end == '\\' && *(end + 1)) { | ||
606 | *out++ = *(end + 1); | ||
607 | end += 2; | ||
608 | continue; | ||
609 | } | ||
610 | |||
611 | if (isspace(*end)) | ||
612 | break; /* end of token */ | ||
613 | |||
614 | *out++ = *end++; | ||
615 | } | ||
616 | |||
617 | /* have we already filled the array ? */ | ||
618 | if ((*argc + 1) > array_size) { | ||
619 | argv = realloc_argv(&array_size, argv); | ||
620 | if (!argv) | ||
621 | return -ENOMEM; | ||
622 | } | ||
623 | |||
624 | /* we know this is whitespace */ | ||
625 | if (*end) | ||
626 | end++; | ||
627 | |||
628 | /* terminate the string and put it in the array */ | ||
629 | *out = '\0'; | ||
630 | argv[*argc] = start; | ||
631 | (*argc)++; | ||
632 | } | ||
633 | |||
634 | *argvp = argv; | ||
635 | return 0; | ||
636 | } | ||
637 | |||
638 | static void check_for_valid_limits(struct io_restrictions *rs) | ||
639 | { | ||
640 | if (!rs->max_sectors) | ||
641 | rs->max_sectors = MAX_SECTORS; | ||
642 | if (!rs->max_phys_segments) | ||
643 | rs->max_phys_segments = MAX_PHYS_SEGMENTS; | ||
644 | if (!rs->max_hw_segments) | ||
645 | rs->max_hw_segments = MAX_HW_SEGMENTS; | ||
646 | if (!rs->hardsect_size) | ||
647 | rs->hardsect_size = 1 << SECTOR_SHIFT; | ||
648 | if (!rs->max_segment_size) | ||
649 | rs->max_segment_size = MAX_SEGMENT_SIZE; | ||
650 | if (!rs->seg_boundary_mask) | ||
651 | rs->seg_boundary_mask = -1; | ||
652 | } | ||
653 | |||
654 | int dm_table_add_target(struct dm_table *t, const char *type, | ||
655 | sector_t start, sector_t len, char *params) | ||
656 | { | ||
657 | int r = -EINVAL, argc; | ||
658 | char **argv; | ||
659 | struct dm_target *tgt; | ||
660 | |||
661 | if ((r = check_space(t))) | ||
662 | return r; | ||
663 | |||
664 | tgt = t->targets + t->num_targets; | ||
665 | memset(tgt, 0, sizeof(*tgt)); | ||
666 | |||
667 | if (!len) { | ||
668 | tgt->error = "zero-length target"; | ||
669 | DMERR("%s", tgt->error); | ||
670 | return -EINVAL; | ||
671 | } | ||
672 | |||
673 | tgt->type = dm_get_target_type(type); | ||
674 | if (!tgt->type) { | ||
675 | tgt->error = "unknown target type"; | ||
676 | DMERR("%s", tgt->error); | ||
677 | return -EINVAL; | ||
678 | } | ||
679 | |||
680 | tgt->table = t; | ||
681 | tgt->begin = start; | ||
682 | tgt->len = len; | ||
683 | tgt->error = "Unknown error"; | ||
684 | |||
685 | /* | ||
686 | * Does this target adjoin the previous one ? | ||
687 | */ | ||
688 | if (!adjoin(t, tgt)) { | ||
689 | tgt->error = "Gap in table"; | ||
690 | r = -EINVAL; | ||
691 | goto bad; | ||
692 | } | ||
693 | |||
694 | r = dm_split_args(&argc, &argv, params); | ||
695 | if (r) { | ||
696 | tgt->error = "couldn't split parameters (insufficient memory)"; | ||
697 | goto bad; | ||
698 | } | ||
699 | |||
700 | r = tgt->type->ctr(tgt, argc, argv); | ||
701 | kfree(argv); | ||
702 | if (r) | ||
703 | goto bad; | ||
704 | |||
705 | t->highs[t->num_targets++] = tgt->begin + tgt->len - 1; | ||
706 | |||
707 | /* FIXME: the plan is to combine high here and then have | ||
708 | * the merge fn apply the target level restrictions. */ | ||
709 | combine_restrictions_low(&t->limits, &tgt->limits); | ||
710 | return 0; | ||
711 | |||
712 | bad: | ||
713 | DMERR("%s", tgt->error); | ||
714 | dm_put_target_type(tgt->type); | ||
715 | return r; | ||
716 | } | ||
717 | |||
718 | static int setup_indexes(struct dm_table *t) | ||
719 | { | ||
720 | int i; | ||
721 | unsigned int total = 0; | ||
722 | sector_t *indexes; | ||
723 | |||
724 | /* allocate the space for *all* the indexes */ | ||
725 | for (i = t->depth - 2; i >= 0; i--) { | ||
726 | t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE); | ||
727 | total += t->counts[i]; | ||
728 | } | ||
729 | |||
730 | indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE); | ||
731 | if (!indexes) | ||
732 | return -ENOMEM; | ||
733 | |||
734 | /* set up internal nodes, bottom-up */ | ||
735 | for (i = t->depth - 2, total = 0; i >= 0; i--) { | ||
736 | t->index[i] = indexes; | ||
737 | indexes += (KEYS_PER_NODE * t->counts[i]); | ||
738 | setup_btree_index(i, t); | ||
739 | } | ||
740 | |||
741 | return 0; | ||
742 | } | ||
743 | |||
744 | /* | ||
745 | * Builds the btree to index the map. | ||
746 | */ | ||
747 | int dm_table_complete(struct dm_table *t) | ||
748 | { | ||
749 | int r = 0; | ||
750 | unsigned int leaf_nodes; | ||
751 | |||
752 | check_for_valid_limits(&t->limits); | ||
753 | |||
754 | /* how many indexes will the btree have ? */ | ||
755 | leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE); | ||
756 | t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE); | ||
757 | |||
758 | /* leaf layer has already been set up */ | ||
759 | t->counts[t->depth - 1] = leaf_nodes; | ||
760 | t->index[t->depth - 1] = t->highs; | ||
761 | |||
762 | if (t->depth >= 2) | ||
763 | r = setup_indexes(t); | ||
764 | |||
765 | return r; | ||
766 | } | ||
767 | |||
768 | static DECLARE_MUTEX(_event_lock); | ||
769 | void dm_table_event_callback(struct dm_table *t, | ||
770 | void (*fn)(void *), void *context) | ||
771 | { | ||
772 | down(&_event_lock); | ||
773 | t->event_fn = fn; | ||
774 | t->event_context = context; | ||
775 | up(&_event_lock); | ||
776 | } | ||
777 | |||
778 | void dm_table_event(struct dm_table *t) | ||
779 | { | ||
780 | /* | ||
781 | * You can no longer call dm_table_event() from interrupt | ||
782 | * context, use a bottom half instead. | ||
783 | */ | ||
784 | BUG_ON(in_interrupt()); | ||
785 | |||
786 | down(&_event_lock); | ||
787 | if (t->event_fn) | ||
788 | t->event_fn(t->event_context); | ||
789 | up(&_event_lock); | ||
790 | } | ||
791 | |||
792 | sector_t dm_table_get_size(struct dm_table *t) | ||
793 | { | ||
794 | return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0; | ||
795 | } | ||
796 | |||
797 | struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index) | ||
798 | { | ||
799 | if (index > t->num_targets) | ||
800 | return NULL; | ||
801 | |||
802 | return t->targets + index; | ||
803 | } | ||
804 | |||
805 | /* | ||
806 | * Search the btree for the correct target. | ||
807 | */ | ||
808 | struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) | ||
809 | { | ||
810 | unsigned int l, n = 0, k = 0; | ||
811 | sector_t *node; | ||
812 | |||
813 | for (l = 0; l < t->depth; l++) { | ||
814 | n = get_child(n, k); | ||
815 | node = get_node(t, l, n); | ||
816 | |||
817 | for (k = 0; k < KEYS_PER_NODE; k++) | ||
818 | if (node[k] >= sector) | ||
819 | break; | ||
820 | } | ||
821 | |||
822 | return &t->targets[(KEYS_PER_NODE * n) + k]; | ||
823 | } | ||
824 | |||
825 | void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q) | ||
826 | { | ||
827 | /* | ||
828 | * Make sure we obey the optimistic sub devices | ||
829 | * restrictions. | ||
830 | */ | ||
831 | blk_queue_max_sectors(q, t->limits.max_sectors); | ||
832 | q->max_phys_segments = t->limits.max_phys_segments; | ||
833 | q->max_hw_segments = t->limits.max_hw_segments; | ||
834 | q->hardsect_size = t->limits.hardsect_size; | ||
835 | q->max_segment_size = t->limits.max_segment_size; | ||
836 | q->seg_boundary_mask = t->limits.seg_boundary_mask; | ||
837 | } | ||
838 | |||
839 | unsigned int dm_table_get_num_targets(struct dm_table *t) | ||
840 | { | ||
841 | return t->num_targets; | ||
842 | } | ||
843 | |||
844 | struct list_head *dm_table_get_devices(struct dm_table *t) | ||
845 | { | ||
846 | return &t->devices; | ||
847 | } | ||
848 | |||
849 | int dm_table_get_mode(struct dm_table *t) | ||
850 | { | ||
851 | return t->mode; | ||
852 | } | ||
853 | |||
854 | static void suspend_targets(struct dm_table *t, unsigned postsuspend) | ||
855 | { | ||
856 | int i = t->num_targets; | ||
857 | struct dm_target *ti = t->targets; | ||
858 | |||
859 | while (i--) { | ||
860 | if (postsuspend) { | ||
861 | if (ti->type->postsuspend) | ||
862 | ti->type->postsuspend(ti); | ||
863 | } else if (ti->type->presuspend) | ||
864 | ti->type->presuspend(ti); | ||
865 | |||
866 | ti++; | ||
867 | } | ||
868 | } | ||
869 | |||
870 | void dm_table_presuspend_targets(struct dm_table *t) | ||
871 | { | ||
872 | return suspend_targets(t, 0); | ||
873 | } | ||
874 | |||
875 | void dm_table_postsuspend_targets(struct dm_table *t) | ||
876 | { | ||
877 | return suspend_targets(t, 1); | ||
878 | } | ||
879 | |||
880 | void dm_table_resume_targets(struct dm_table *t) | ||
881 | { | ||
882 | int i; | ||
883 | |||
884 | for (i = 0; i < t->num_targets; i++) { | ||
885 | struct dm_target *ti = t->targets + i; | ||
886 | |||
887 | if (ti->type->resume) | ||
888 | ti->type->resume(ti); | ||
889 | } | ||
890 | } | ||
891 | |||
892 | int dm_table_any_congested(struct dm_table *t, int bdi_bits) | ||
893 | { | ||
894 | struct list_head *d, *devices; | ||
895 | int r = 0; | ||
896 | |||
897 | devices = dm_table_get_devices(t); | ||
898 | for (d = devices->next; d != devices; d = d->next) { | ||
899 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | ||
900 | request_queue_t *q = bdev_get_queue(dd->bdev); | ||
901 | r |= bdi_congested(&q->backing_dev_info, bdi_bits); | ||
902 | } | ||
903 | |||
904 | return r; | ||
905 | } | ||
906 | |||
907 | void dm_table_unplug_all(struct dm_table *t) | ||
908 | { | ||
909 | struct list_head *d, *devices = dm_table_get_devices(t); | ||
910 | |||
911 | for (d = devices->next; d != devices; d = d->next) { | ||
912 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | ||
913 | request_queue_t *q = bdev_get_queue(dd->bdev); | ||
914 | |||
915 | if (q->unplug_fn) | ||
916 | q->unplug_fn(q); | ||
917 | } | ||
918 | } | ||
919 | |||
920 | int dm_table_flush_all(struct dm_table *t) | ||
921 | { | ||
922 | struct list_head *d, *devices = dm_table_get_devices(t); | ||
923 | int ret = 0; | ||
924 | |||
925 | for (d = devices->next; d != devices; d = d->next) { | ||
926 | struct dm_dev *dd = list_entry(d, struct dm_dev, list); | ||
927 | request_queue_t *q = bdev_get_queue(dd->bdev); | ||
928 | int err; | ||
929 | |||
930 | if (!q->issue_flush_fn) | ||
931 | err = -EOPNOTSUPP; | ||
932 | else | ||
933 | err = q->issue_flush_fn(q, dd->bdev->bd_disk, NULL); | ||
934 | |||
935 | if (!ret) | ||
936 | ret = err; | ||
937 | } | ||
938 | |||
939 | return ret; | ||
940 | } | ||
941 | |||
942 | EXPORT_SYMBOL(dm_vcalloc); | ||
943 | EXPORT_SYMBOL(dm_get_device); | ||
944 | EXPORT_SYMBOL(dm_put_device); | ||
945 | EXPORT_SYMBOL(dm_table_event); | ||
946 | EXPORT_SYMBOL(dm_table_get_mode); | ||
947 | EXPORT_SYMBOL(dm_table_put); | ||
948 | EXPORT_SYMBOL(dm_table_get); | ||
949 | EXPORT_SYMBOL(dm_table_unplug_all); | ||
950 | EXPORT_SYMBOL(dm_table_flush_all); | ||