blob: 05befa91807a332b87676bef2cc47de92417e1eb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Arjan van de Ven48c9c272006-03-27 01:18:20 -080017#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <asm/atomic.h>
19
Alasdair G Kergon72d94862006-06-26 00:27:35 -070020#define DM_MSG_PREFIX "table"
21
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#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
27struct dm_table {
Mike Anderson1134e5a2006-03-27 01:17:54 -080028 struct mapped_device *md;
Linus Torvalds1da177e2005-04-16 15:20:36 -070029 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 */
65static 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 */
85static 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
102 lhs->seg_boundary_mask =
103 min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask);
NeilBrown969429b2006-03-27 01:17:49 -0800104
105 lhs->no_cluster |= rhs->no_cluster;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106}
107
108/*
109 * Calculate the index of the child node of the n'th node k'th key.
110 */
111static inline unsigned int get_child(unsigned int n, unsigned int k)
112{
113 return (n * CHILDREN_PER_NODE) + k;
114}
115
116/*
117 * Return the n'th node of level l from table t.
118 */
119static inline sector_t *get_node(struct dm_table *t,
120 unsigned int l, unsigned int n)
121{
122 return t->index[l] + (n * KEYS_PER_NODE);
123}
124
125/*
126 * Return the highest key that you could lookup from the n'th
127 * node on level l of the btree.
128 */
129static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
130{
131 for (; l < t->depth - 1; l++)
132 n = get_child(n, CHILDREN_PER_NODE - 1);
133
134 if (n >= t->counts[l])
135 return (sector_t) - 1;
136
137 return get_node(t, l, n)[KEYS_PER_NODE - 1];
138}
139
140/*
141 * Fills in a level of the btree based on the highs of the level
142 * below it.
143 */
144static int setup_btree_index(unsigned int l, struct dm_table *t)
145{
146 unsigned int n, k;
147 sector_t *node;
148
149 for (n = 0U; n < t->counts[l]; n++) {
150 node = get_node(t, l, n);
151
152 for (k = 0U; k < KEYS_PER_NODE; k++)
153 node[k] = high(t, l + 1, get_child(n, k));
154 }
155
156 return 0;
157}
158
159void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
160{
161 unsigned long size;
162 void *addr;
163
164 /*
165 * Check that we're not going to overflow.
166 */
167 if (nmemb > (ULONG_MAX / elem_size))
168 return NULL;
169
170 size = nmemb * elem_size;
171 addr = vmalloc(size);
172 if (addr)
173 memset(addr, 0, size);
174
175 return addr;
176}
177
178/*
179 * highs, and targets are managed as dynamic arrays during a
180 * table load.
181 */
182static int alloc_targets(struct dm_table *t, unsigned int num)
183{
184 sector_t *n_highs;
185 struct dm_target *n_targets;
186 int n = t->num_targets;
187
188 /*
189 * Allocate both the target array and offset array at once.
190 */
191 n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) +
192 sizeof(sector_t));
193 if (!n_highs)
194 return -ENOMEM;
195
196 n_targets = (struct dm_target *) (n_highs + num);
197
198 if (n) {
199 memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
200 memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
201 }
202
203 memset(n_highs + n, -1, sizeof(*n_highs) * (num - n));
204 vfree(t->highs);
205
206 t->num_allocated = num;
207 t->highs = n_highs;
208 t->targets = n_targets;
209
210 return 0;
211}
212
Mike Anderson1134e5a2006-03-27 01:17:54 -0800213int dm_table_create(struct dm_table **result, int mode,
214 unsigned num_targets, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 struct dm_table *t = kmalloc(sizeof(*t), GFP_KERNEL);
217
218 if (!t)
219 return -ENOMEM;
220
221 memset(t, 0, sizeof(*t));
222 INIT_LIST_HEAD(&t->devices);
223 atomic_set(&t->holders, 1);
224
225 if (!num_targets)
226 num_targets = KEYS_PER_NODE;
227
228 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
229
230 if (alloc_targets(t, num_targets)) {
231 kfree(t);
232 t = NULL;
233 return -ENOMEM;
234 }
235
236 t->mode = mode;
Mike Anderson1134e5a2006-03-27 01:17:54 -0800237 t->md = md;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 *result = t;
239 return 0;
240}
241
David Teiglandc2ade422006-06-26 00:27:33 -0700242int dm_create_error_table(struct dm_table **result, struct mapped_device *md)
243{
244 struct dm_table *t;
245 sector_t dev_size = 1;
246 int r;
247
248 /*
249 * Find current size of device.
250 * Default to 1 sector if inactive.
251 */
252 t = dm_get_table(md);
253 if (t) {
254 dev_size = dm_table_get_size(t);
255 dm_table_put(t);
256 }
257
258 r = dm_table_create(&t, FMODE_READ, 1, md);
259 if (r)
260 return r;
261
262 r = dm_table_add_target(t, "error", 0, dev_size, NULL);
263 if (r)
264 goto out;
265
266 r = dm_table_complete(t);
267 if (r)
268 goto out;
269
270 *result = t;
271
272out:
273 if (r)
274 dm_table_put(t);
275
276 return r;
277}
278EXPORT_SYMBOL_GPL(dm_create_error_table);
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280static void free_devices(struct list_head *devices)
281{
282 struct list_head *tmp, *next;
283
284 for (tmp = devices->next; tmp != devices; tmp = next) {
285 struct dm_dev *dd = list_entry(tmp, struct dm_dev, list);
286 next = tmp->next;
287 kfree(dd);
288 }
289}
290
Alasdair G Kergon5e198d92005-05-05 16:16:09 -0700291static void table_destroy(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
293 unsigned int i;
294
295 /* free the indexes (see dm_table_complete) */
296 if (t->depth >= 2)
297 vfree(t->index[t->depth - 2]);
298
299 /* free the targets */
300 for (i = 0; i < t->num_targets; i++) {
301 struct dm_target *tgt = t->targets + i;
302
303 if (tgt->type->dtr)
304 tgt->type->dtr(tgt);
305
306 dm_put_target_type(tgt->type);
307 }
308
309 vfree(t->highs);
310
311 /* free the device list */
312 if (t->devices.next != &t->devices) {
313 DMWARN("devices still present during destroy: "
314 "dm_table_remove_device calls missing");
315
316 free_devices(&t->devices);
317 }
318
319 kfree(t);
320}
321
322void dm_table_get(struct dm_table *t)
323{
324 atomic_inc(&t->holders);
325}
326
327void dm_table_put(struct dm_table *t)
328{
329 if (!t)
330 return;
331
332 if (atomic_dec_and_test(&t->holders))
333 table_destroy(t);
334}
335
336/*
337 * Checks to see if we need to extend highs or targets.
338 */
339static inline int check_space(struct dm_table *t)
340{
341 if (t->num_targets >= t->num_allocated)
342 return alloc_targets(t, t->num_allocated * 2);
343
344 return 0;
345}
346
347/*
348 * Convert a device path to a dev_t.
349 */
350static int lookup_device(const char *path, dev_t *dev)
351{
352 int r;
353 struct nameidata nd;
354 struct inode *inode;
355
356 if ((r = path_lookup(path, LOOKUP_FOLLOW, &nd)))
357 return r;
358
359 inode = nd.dentry->d_inode;
360 if (!inode) {
361 r = -ENOENT;
362 goto out;
363 }
364
365 if (!S_ISBLK(inode->i_mode)) {
366 r = -ENOTBLK;
367 goto out;
368 }
369
370 *dev = inode->i_rdev;
371
372 out:
373 path_release(&nd);
374 return r;
375}
376
377/*
378 * See if we've already got a device in the list.
379 */
380static struct dm_dev *find_device(struct list_head *l, dev_t dev)
381{
382 struct dm_dev *dd;
383
384 list_for_each_entry (dd, l, list)
385 if (dd->bdev->bd_dev == dev)
386 return dd;
387
388 return NULL;
389}
390
391/*
392 * Open a device so we can use it as a map destination.
393 */
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800394static int open_dev(struct dm_dev *d, dev_t dev, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395{
396 static char *_claim_ptr = "I belong to device-mapper";
397 struct block_device *bdev;
398
399 int r;
400
Eric Sesterhenn547bc922006-03-26 18:22:50 +0200401 BUG_ON(d->bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
403 bdev = open_by_devnum(dev, d->mode);
404 if (IS_ERR(bdev))
405 return PTR_ERR(bdev);
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800406 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 if (r)
408 blkdev_put(bdev);
409 else
410 d->bdev = bdev;
411 return r;
412}
413
414/*
415 * Close a device that we've been using.
416 */
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800417static void close_dev(struct dm_dev *d, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418{
419 if (!d->bdev)
420 return;
421
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800422 bd_release_from_disk(d->bdev, dm_disk(md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 blkdev_put(d->bdev);
424 d->bdev = NULL;
425}
426
427/*
428 * If possible (ie. blk_size[major] is set), this checks an area
429 * of a destination device is valid.
430 */
431static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len)
432{
433 sector_t dev_size;
434 dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT;
435 return ((start < dev_size) && (len <= (dev_size - start)));
436}
437
438/*
439 * This upgrades the mode on an already open dm_dev. Being
440 * careful to leave things as they were if we fail to reopen the
441 * device.
442 */
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800443static int upgrade_mode(struct dm_dev *dd, int new_mode, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
445 int r;
446 struct dm_dev dd_copy;
447 dev_t dev = dd->bdev->bd_dev;
448
449 dd_copy = *dd;
450
451 dd->mode |= new_mode;
452 dd->bdev = NULL;
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800453 r = open_dev(dd, dev, md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 if (!r)
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800455 close_dev(&dd_copy, md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 else
457 *dd = dd_copy;
458
459 return r;
460}
461
462/*
463 * Add a device to the list, or just increment the usage count if
464 * it's already present.
465 */
466static int __table_get_device(struct dm_table *t, struct dm_target *ti,
467 const char *path, sector_t start, sector_t len,
468 int mode, struct dm_dev **result)
469{
470 int r;
471 dev_t dev;
472 struct dm_dev *dd;
473 unsigned int major, minor;
474
Eric Sesterhenn547bc922006-03-26 18:22:50 +0200475 BUG_ON(!t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477 if (sscanf(path, "%u:%u", &major, &minor) == 2) {
478 /* Extract the major/minor numbers */
479 dev = MKDEV(major, minor);
480 if (MAJOR(dev) != major || MINOR(dev) != minor)
481 return -EOVERFLOW;
482 } else {
483 /* convert the path to a device */
484 if ((r = lookup_device(path, &dev)))
485 return r;
486 }
487
488 dd = find_device(&t->devices, dev);
489 if (!dd) {
490 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
491 if (!dd)
492 return -ENOMEM;
493
494 dd->mode = mode;
495 dd->bdev = NULL;
496
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800497 if ((r = open_dev(dd, dev, t->md))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 kfree(dd);
499 return r;
500 }
501
502 format_dev_t(dd->name, dev);
503
504 atomic_set(&dd->count, 0);
505 list_add(&dd->list, &t->devices);
506
507 } else if (dd->mode != (mode | dd->mode)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800508 r = upgrade_mode(dd, mode, t->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 if (r)
510 return r;
511 }
512 atomic_inc(&dd->count);
513
514 if (!check_device_area(dd, start, len)) {
515 DMWARN("device %s too small for target", path);
516 dm_put_device(ti, dd);
517 return -EINVAL;
518 }
519
520 *result = dd;
521
522 return 0;
523}
524
Bryn Reeves3cb40212006-10-03 01:15:42 -0700525void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
526{
527 request_queue_t *q = bdev_get_queue(bdev);
528 struct io_restrictions *rs = &ti->limits;
529
530 /*
531 * Combine the device limits low.
532 *
533 * FIXME: if we move an io_restriction struct
534 * into q this would just be a call to
535 * combine_restrictions_low()
536 */
537 rs->max_sectors =
538 min_not_zero(rs->max_sectors, q->max_sectors);
539
540 /* FIXME: Device-Mapper on top of RAID-0 breaks because DM
541 * currently doesn't honor MD's merge_bvec_fn routine.
542 * In this case, we'll force DM to use PAGE_SIZE or
543 * smaller I/O, just to be safe. A better fix is in the
544 * works, but add this for the time being so it will at
545 * least operate correctly.
546 */
547 if (q->merge_bvec_fn)
548 rs->max_sectors =
549 min_not_zero(rs->max_sectors,
550 (unsigned int) (PAGE_SIZE >> 9));
551
552 rs->max_phys_segments =
553 min_not_zero(rs->max_phys_segments,
554 q->max_phys_segments);
555
556 rs->max_hw_segments =
557 min_not_zero(rs->max_hw_segments, q->max_hw_segments);
558
559 rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size);
560
561 rs->max_segment_size =
562 min_not_zero(rs->max_segment_size, q->max_segment_size);
563
564 rs->seg_boundary_mask =
565 min_not_zero(rs->seg_boundary_mask,
566 q->seg_boundary_mask);
567
568 rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
569}
570EXPORT_SYMBOL_GPL(dm_set_device_limits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572int dm_get_device(struct dm_target *ti, const char *path, sector_t start,
573 sector_t len, int mode, struct dm_dev **result)
574{
575 int r = __table_get_device(ti->table, ti, path,
576 start, len, mode, result);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577
Bryn Reeves3cb40212006-10-03 01:15:42 -0700578 if (!r)
579 dm_set_device_limits(ti, (*result)->bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
581 return r;
582}
583
584/*
585 * Decrement a devices use count and remove it if necessary.
586 */
587void dm_put_device(struct dm_target *ti, struct dm_dev *dd)
588{
589 if (atomic_dec_and_test(&dd->count)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800590 close_dev(dd, ti->table->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 list_del(&dd->list);
592 kfree(dd);
593 }
594}
595
596/*
597 * Checks to see if the target joins onto the end of the table.
598 */
599static int adjoin(struct dm_table *table, struct dm_target *ti)
600{
601 struct dm_target *prev;
602
603 if (!table->num_targets)
604 return !ti->begin;
605
606 prev = &table->targets[table->num_targets - 1];
607 return (ti->begin == (prev->begin + prev->len));
608}
609
610/*
611 * Used to dynamically allocate the arg array.
612 */
613static char **realloc_argv(unsigned *array_size, char **old_argv)
614{
615 char **argv;
616 unsigned new_size;
617
618 new_size = *array_size ? *array_size * 2 : 64;
619 argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
620 if (argv) {
621 memcpy(argv, old_argv, *array_size * sizeof(*argv));
622 *array_size = new_size;
623 }
624
625 kfree(old_argv);
626 return argv;
627}
628
629/*
630 * Destructively splits up the argument list to pass to ctr.
631 */
632int dm_split_args(int *argc, char ***argvp, char *input)
633{
634 char *start, *end = input, *out, **argv = NULL;
635 unsigned array_size = 0;
636
637 *argc = 0;
David Teigland814d6862006-06-26 00:27:31 -0700638
639 if (!input) {
640 *argvp = NULL;
641 return 0;
642 }
643
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 argv = realloc_argv(&array_size, argv);
645 if (!argv)
646 return -ENOMEM;
647
648 while (1) {
649 start = end;
650
651 /* Skip whitespace */
652 while (*start && isspace(*start))
653 start++;
654
655 if (!*start)
656 break; /* success, we hit the end */
657
658 /* 'out' is used to remove any back-quotes */
659 end = out = start;
660 while (*end) {
661 /* Everything apart from '\0' can be quoted */
662 if (*end == '\\' && *(end + 1)) {
663 *out++ = *(end + 1);
664 end += 2;
665 continue;
666 }
667
668 if (isspace(*end))
669 break; /* end of token */
670
671 *out++ = *end++;
672 }
673
674 /* have we already filled the array ? */
675 if ((*argc + 1) > array_size) {
676 argv = realloc_argv(&array_size, argv);
677 if (!argv)
678 return -ENOMEM;
679 }
680
681 /* we know this is whitespace */
682 if (*end)
683 end++;
684
685 /* terminate the string and put it in the array */
686 *out = '\0';
687 argv[*argc] = start;
688 (*argc)++;
689 }
690
691 *argvp = argv;
692 return 0;
693}
694
695static void check_for_valid_limits(struct io_restrictions *rs)
696{
697 if (!rs->max_sectors)
Mike Christiedefd94b2005-12-05 02:37:06 -0600698 rs->max_sectors = SAFE_MAX_SECTORS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 if (!rs->max_phys_segments)
700 rs->max_phys_segments = MAX_PHYS_SEGMENTS;
701 if (!rs->max_hw_segments)
702 rs->max_hw_segments = MAX_HW_SEGMENTS;
703 if (!rs->hardsect_size)
704 rs->hardsect_size = 1 << SECTOR_SHIFT;
705 if (!rs->max_segment_size)
706 rs->max_segment_size = MAX_SEGMENT_SIZE;
707 if (!rs->seg_boundary_mask)
708 rs->seg_boundary_mask = -1;
709}
710
711int dm_table_add_target(struct dm_table *t, const char *type,
712 sector_t start, sector_t len, char *params)
713{
714 int r = -EINVAL, argc;
715 char **argv;
716 struct dm_target *tgt;
717
718 if ((r = check_space(t)))
719 return r;
720
721 tgt = t->targets + t->num_targets;
722 memset(tgt, 0, sizeof(*tgt));
723
724 if (!len) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700725 DMERR("%s: zero-length target", dm_device_name(t->md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 return -EINVAL;
727 }
728
729 tgt->type = dm_get_target_type(type);
730 if (!tgt->type) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700731 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
732 type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 return -EINVAL;
734 }
735
736 tgt->table = t;
737 tgt->begin = start;
738 tgt->len = len;
739 tgt->error = "Unknown error";
740
741 /*
742 * Does this target adjoin the previous one ?
743 */
744 if (!adjoin(t, tgt)) {
745 tgt->error = "Gap in table";
746 r = -EINVAL;
747 goto bad;
748 }
749
750 r = dm_split_args(&argc, &argv, params);
751 if (r) {
752 tgt->error = "couldn't split parameters (insufficient memory)";
753 goto bad;
754 }
755
756 r = tgt->type->ctr(tgt, argc, argv);
757 kfree(argv);
758 if (r)
759 goto bad;
760
761 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
762
763 /* FIXME: the plan is to combine high here and then have
764 * the merge fn apply the target level restrictions. */
765 combine_restrictions_low(&t->limits, &tgt->limits);
766 return 0;
767
768 bad:
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700769 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 dm_put_target_type(tgt->type);
771 return r;
772}
773
774static int setup_indexes(struct dm_table *t)
775{
776 int i;
777 unsigned int total = 0;
778 sector_t *indexes;
779
780 /* allocate the space for *all* the indexes */
781 for (i = t->depth - 2; i >= 0; i--) {
782 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
783 total += t->counts[i];
784 }
785
786 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
787 if (!indexes)
788 return -ENOMEM;
789
790 /* set up internal nodes, bottom-up */
791 for (i = t->depth - 2, total = 0; i >= 0; i--) {
792 t->index[i] = indexes;
793 indexes += (KEYS_PER_NODE * t->counts[i]);
794 setup_btree_index(i, t);
795 }
796
797 return 0;
798}
799
800/*
801 * Builds the btree to index the map.
802 */
803int dm_table_complete(struct dm_table *t)
804{
805 int r = 0;
806 unsigned int leaf_nodes;
807
808 check_for_valid_limits(&t->limits);
809
810 /* how many indexes will the btree have ? */
811 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
812 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
813
814 /* leaf layer has already been set up */
815 t->counts[t->depth - 1] = leaf_nodes;
816 t->index[t->depth - 1] = t->highs;
817
818 if (t->depth >= 2)
819 r = setup_indexes(t);
820
821 return r;
822}
823
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800824static DEFINE_MUTEX(_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825void dm_table_event_callback(struct dm_table *t,
826 void (*fn)(void *), void *context)
827{
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800828 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 t->event_fn = fn;
830 t->event_context = context;
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800831 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832}
833
834void dm_table_event(struct dm_table *t)
835{
836 /*
837 * You can no longer call dm_table_event() from interrupt
838 * context, use a bottom half instead.
839 */
840 BUG_ON(in_interrupt());
841
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800842 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 if (t->event_fn)
844 t->event_fn(t->event_context);
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800845 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846}
847
848sector_t dm_table_get_size(struct dm_table *t)
849{
850 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
851}
852
853struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
854{
Milan Broz14353532006-06-26 00:27:27 -0700855 if (index >= t->num_targets)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 return NULL;
857
858 return t->targets + index;
859}
860
861/*
862 * Search the btree for the correct target.
863 */
864struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
865{
866 unsigned int l, n = 0, k = 0;
867 sector_t *node;
868
869 for (l = 0; l < t->depth; l++) {
870 n = get_child(n, k);
871 node = get_node(t, l, n);
872
873 for (k = 0; k < KEYS_PER_NODE; k++)
874 if (node[k] >= sector)
875 break;
876 }
877
878 return &t->targets[(KEYS_PER_NODE * n) + k];
879}
880
881void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q)
882{
883 /*
884 * Make sure we obey the optimistic sub devices
885 * restrictions.
886 */
887 blk_queue_max_sectors(q, t->limits.max_sectors);
888 q->max_phys_segments = t->limits.max_phys_segments;
889 q->max_hw_segments = t->limits.max_hw_segments;
890 q->hardsect_size = t->limits.hardsect_size;
891 q->max_segment_size = t->limits.max_segment_size;
892 q->seg_boundary_mask = t->limits.seg_boundary_mask;
NeilBrown969429b2006-03-27 01:17:49 -0800893 if (t->limits.no_cluster)
894 q->queue_flags &= ~(1 << QUEUE_FLAG_CLUSTER);
895 else
896 q->queue_flags |= (1 << QUEUE_FLAG_CLUSTER);
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
899
900unsigned int dm_table_get_num_targets(struct dm_table *t)
901{
902 return t->num_targets;
903}
904
905struct list_head *dm_table_get_devices(struct dm_table *t)
906{
907 return &t->devices;
908}
909
910int dm_table_get_mode(struct dm_table *t)
911{
912 return t->mode;
913}
914
915static void suspend_targets(struct dm_table *t, unsigned postsuspend)
916{
917 int i = t->num_targets;
918 struct dm_target *ti = t->targets;
919
920 while (i--) {
921 if (postsuspend) {
922 if (ti->type->postsuspend)
923 ti->type->postsuspend(ti);
924 } else if (ti->type->presuspend)
925 ti->type->presuspend(ti);
926
927 ti++;
928 }
929}
930
931void dm_table_presuspend_targets(struct dm_table *t)
932{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -0700933 if (!t)
934 return;
935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 return suspend_targets(t, 0);
937}
938
939void dm_table_postsuspend_targets(struct dm_table *t)
940{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -0700941 if (!t)
942 return;
943
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 return suspend_targets(t, 1);
945}
946
Milan Broz8757b772006-10-03 01:15:36 -0700947int dm_table_resume_targets(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948{
Milan Broz8757b772006-10-03 01:15:36 -0700949 int i, r = 0;
950
951 for (i = 0; i < t->num_targets; i++) {
952 struct dm_target *ti = t->targets + i;
953
954 if (!ti->type->preresume)
955 continue;
956
957 r = ti->type->preresume(ti);
958 if (r)
959 return r;
960 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
962 for (i = 0; i < t->num_targets; i++) {
963 struct dm_target *ti = t->targets + i;
964
965 if (ti->type->resume)
966 ti->type->resume(ti);
967 }
Milan Broz8757b772006-10-03 01:15:36 -0700968
969 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970}
971
972int dm_table_any_congested(struct dm_table *t, int bdi_bits)
973{
974 struct list_head *d, *devices;
975 int r = 0;
976
977 devices = dm_table_get_devices(t);
978 for (d = devices->next; d != devices; d = d->next) {
979 struct dm_dev *dd = list_entry(d, struct dm_dev, list);
980 request_queue_t *q = bdev_get_queue(dd->bdev);
981 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
982 }
983
984 return r;
985}
986
987void dm_table_unplug_all(struct dm_table *t)
988{
989 struct list_head *d, *devices = dm_table_get_devices(t);
990
991 for (d = devices->next; d != devices; d = d->next) {
992 struct dm_dev *dd = list_entry(d, struct dm_dev, list);
993 request_queue_t *q = bdev_get_queue(dd->bdev);
994
995 if (q->unplug_fn)
996 q->unplug_fn(q);
997 }
998}
999
1000int dm_table_flush_all(struct dm_table *t)
1001{
1002 struct list_head *d, *devices = dm_table_get_devices(t);
1003 int ret = 0;
Bryn Reeves999d8162006-10-03 01:15:43 -07001004 unsigned i;
1005
1006 for (i = 0; i < t->num_targets; i++)
1007 if (t->targets[i].type->flush)
1008 t->targets[i].type->flush(&t->targets[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
1010 for (d = devices->next; d != devices; d = d->next) {
1011 struct dm_dev *dd = list_entry(d, struct dm_dev, list);
1012 request_queue_t *q = bdev_get_queue(dd->bdev);
1013 int err;
1014
1015 if (!q->issue_flush_fn)
1016 err = -EOPNOTSUPP;
1017 else
1018 err = q->issue_flush_fn(q, dd->bdev->bd_disk, NULL);
1019
1020 if (!ret)
1021 ret = err;
1022 }
1023
1024 return ret;
1025}
1026
Mike Anderson1134e5a2006-03-27 01:17:54 -08001027struct mapped_device *dm_table_get_md(struct dm_table *t)
1028{
1029 dm_get(t->md);
1030
1031 return t->md;
1032}
1033
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034EXPORT_SYMBOL(dm_vcalloc);
1035EXPORT_SYMBOL(dm_get_device);
1036EXPORT_SYMBOL(dm_put_device);
1037EXPORT_SYMBOL(dm_table_event);
Alasdair G Kergond5e404c2005-07-12 15:53:05 -07001038EXPORT_SYMBOL(dm_table_get_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039EXPORT_SYMBOL(dm_table_get_mode);
Mike Anderson1134e5a2006-03-27 01:17:54 -08001040EXPORT_SYMBOL(dm_table_get_md);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041EXPORT_SYMBOL(dm_table_put);
1042EXPORT_SYMBOL(dm_table_get);
1043EXPORT_SYMBOL(dm_table_unplug_all);
1044EXPORT_SYMBOL(dm_table_flush_all);