dm: linear add merge
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / drivers / md / dm-table.c
CommitLineData
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
27struct 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 */
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
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 */
116static 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 */
124static 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 */
134static 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 */
149static 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
164void *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 */
187static 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
220int 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
248static void free_devices(struct list_head *devices)
249{
250 struct list_head *tmp, *next;
251
afb24528 252 list_for_each_safe(tmp, next, devices) {
1da177e4 253 struct dm_dev *dd = list_entry(tmp, struct dm_dev, list);
1da177e4
LT
254 kfree(dd);
255 }
256}
257
5e198d94 258static void table_destroy(struct dm_table *t)
1da177e4
LT
259{
260 unsigned int i;
261
262 /* free the indexes (see dm_table_complete) */
263 if (t->depth >= 2)
264 vfree(t->index[t->depth - 2]);
265
266 /* free the targets */
267 for (i = 0; i < t->num_targets; i++) {
268 struct dm_target *tgt = t->targets + i;
269
270 if (tgt->type->dtr)
271 tgt->type->dtr(tgt);
272
273 dm_put_target_type(tgt->type);
274 }
275
276 vfree(t->highs);
277
278 /* free the device list */
279 if (t->devices.next != &t->devices) {
280 DMWARN("devices still present during destroy: "
281 "dm_table_remove_device calls missing");
282
283 free_devices(&t->devices);
284 }
285
286 kfree(t);
287}
288
289void dm_table_get(struct dm_table *t)
290{
291 atomic_inc(&t->holders);
292}
293
294void dm_table_put(struct dm_table *t)
295{
296 if (!t)
297 return;
298
299 if (atomic_dec_and_test(&t->holders))
300 table_destroy(t);
301}
302
303/*
304 * Checks to see if we need to extend highs or targets.
305 */
306static inline int check_space(struct dm_table *t)
307{
308 if (t->num_targets >= t->num_allocated)
309 return alloc_targets(t, t->num_allocated * 2);
310
311 return 0;
312}
313
314/*
315 * Convert a device path to a dev_t.
316 */
317static int lookup_device(const char *path, dev_t *dev)
318{
319 int r;
320 struct nameidata nd;
321 struct inode *inode;
322
323 if ((r = path_lookup(path, LOOKUP_FOLLOW, &nd)))
324 return r;
325
4ac91378 326 inode = nd.path.dentry->d_inode;
1da177e4
LT
327 if (!inode) {
328 r = -ENOENT;
329 goto out;
330 }
331
332 if (!S_ISBLK(inode->i_mode)) {
333 r = -ENOTBLK;
334 goto out;
335 }
336
337 *dev = inode->i_rdev;
338
339 out:
1d957f9b 340 path_put(&nd.path);
1da177e4
LT
341 return r;
342}
343
344/*
345 * See if we've already got a device in the list.
346 */
347static struct dm_dev *find_device(struct list_head *l, dev_t dev)
348{
349 struct dm_dev *dd;
350
351 list_for_each_entry (dd, l, list)
352 if (dd->bdev->bd_dev == dev)
353 return dd;
354
355 return NULL;
356}
357
358/*
359 * Open a device so we can use it as a map destination.
360 */
f165921d 361static int open_dev(struct dm_dev *d, dev_t dev, struct mapped_device *md)
1da177e4
LT
362{
363 static char *_claim_ptr = "I belong to device-mapper";
364 struct block_device *bdev;
365
366 int r;
367
547bc926 368 BUG_ON(d->bdev);
1da177e4
LT
369
370 bdev = open_by_devnum(dev, d->mode);
371 if (IS_ERR(bdev))
372 return PTR_ERR(bdev);
f165921d 373 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
1da177e4
LT
374 if (r)
375 blkdev_put(bdev);
376 else
377 d->bdev = bdev;
378 return r;
379}
380
381/*
382 * Close a device that we've been using.
383 */
f165921d 384static void close_dev(struct dm_dev *d, struct mapped_device *md)
1da177e4
LT
385{
386 if (!d->bdev)
387 return;
388
f165921d 389 bd_release_from_disk(d->bdev, dm_disk(md));
1da177e4
LT
390 blkdev_put(d->bdev);
391 d->bdev = NULL;
392}
393
394/*
2cd54d9b 395 * If possible, this checks an area of a destination device is valid.
1da177e4
LT
396 */
397static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len)
398{
2cd54d9b
MA
399 sector_t dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT;
400
401 if (!dev_size)
402 return 1;
403
1da177e4
LT
404 return ((start < dev_size) && (len <= (dev_size - start)));
405}
406
407/*
408 * This upgrades the mode on an already open dm_dev. Being
409 * careful to leave things as they were if we fail to reopen the
410 * device.
411 */
f165921d 412static int upgrade_mode(struct dm_dev *dd, int new_mode, struct mapped_device *md)
1da177e4
LT
413{
414 int r;
415 struct dm_dev dd_copy;
416 dev_t dev = dd->bdev->bd_dev;
417
418 dd_copy = *dd;
419
420 dd->mode |= new_mode;
421 dd->bdev = NULL;
f165921d 422 r = open_dev(dd, dev, md);
1da177e4 423 if (!r)
f165921d 424 close_dev(&dd_copy, md);
1da177e4
LT
425 else
426 *dd = dd_copy;
427
428 return r;
429}
430
431/*
432 * Add a device to the list, or just increment the usage count if
433 * it's already present.
434 */
435static int __table_get_device(struct dm_table *t, struct dm_target *ti,
436 const char *path, sector_t start, sector_t len,
437 int mode, struct dm_dev **result)
438{
439 int r;
69a2ce72 440 dev_t uninitialized_var(dev);
1da177e4
LT
441 struct dm_dev *dd;
442 unsigned int major, minor;
443
547bc926 444 BUG_ON(!t);
1da177e4
LT
445
446 if (sscanf(path, "%u:%u", &major, &minor) == 2) {
447 /* Extract the major/minor numbers */
448 dev = MKDEV(major, minor);
449 if (MAJOR(dev) != major || MINOR(dev) != minor)
450 return -EOVERFLOW;
451 } else {
452 /* convert the path to a device */
453 if ((r = lookup_device(path, &dev)))
454 return r;
455 }
456
457 dd = find_device(&t->devices, dev);
458 if (!dd) {
459 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
460 if (!dd)
461 return -ENOMEM;
462
463 dd->mode = mode;
464 dd->bdev = NULL;
465
f165921d 466 if ((r = open_dev(dd, dev, t->md))) {
1da177e4
LT
467 kfree(dd);
468 return r;
469 }
470
471 format_dev_t(dd->name, dev);
472
473 atomic_set(&dd->count, 0);
474 list_add(&dd->list, &t->devices);
475
476 } else if (dd->mode != (mode | dd->mode)) {
f165921d 477 r = upgrade_mode(dd, mode, t->md);
1da177e4
LT
478 if (r)
479 return r;
480 }
481 atomic_inc(&dd->count);
482
483 if (!check_device_area(dd, start, len)) {
484 DMWARN("device %s too small for target", path);
485 dm_put_device(ti, dd);
486 return -EINVAL;
487 }
488
489 *result = dd;
490
491 return 0;
492}
493
3cb40214 494void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
1da177e4 495{
165125e1 496 struct request_queue *q = bdev_get_queue(bdev);
3cb40214
BR
497 struct io_restrictions *rs = &ti->limits;
498
499 /*
500 * Combine the device limits low.
501 *
502 * FIXME: if we move an io_restriction struct
503 * into q this would just be a call to
504 * combine_restrictions_low()
505 */
506 rs->max_sectors =
507 min_not_zero(rs->max_sectors, q->max_sectors);
508
509 /* FIXME: Device-Mapper on top of RAID-0 breaks because DM
510 * currently doesn't honor MD's merge_bvec_fn routine.
511 * In this case, we'll force DM to use PAGE_SIZE or
512 * smaller I/O, just to be safe. A better fix is in the
513 * works, but add this for the time being so it will at
514 * least operate correctly.
515 */
516 if (q->merge_bvec_fn)
1da177e4 517 rs->max_sectors =
3cb40214
BR
518 min_not_zero(rs->max_sectors,
519 (unsigned int) (PAGE_SIZE >> 9));
1da177e4 520
3cb40214
BR
521 rs->max_phys_segments =
522 min_not_zero(rs->max_phys_segments,
523 q->max_phys_segments);
1da177e4 524
3cb40214
BR
525 rs->max_hw_segments =
526 min_not_zero(rs->max_hw_segments, q->max_hw_segments);
1da177e4 527
3cb40214 528 rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size);
1da177e4 529
3cb40214
BR
530 rs->max_segment_size =
531 min_not_zero(rs->max_segment_size, q->max_segment_size);
1da177e4 532
91212507
NB
533 rs->max_hw_sectors =
534 min_not_zero(rs->max_hw_sectors, q->max_hw_sectors);
535
3cb40214
BR
536 rs->seg_boundary_mask =
537 min_not_zero(rs->seg_boundary_mask,
538 q->seg_boundary_mask);
1da177e4 539
5ec140e6
VA
540 rs->bounce_pfn = min_not_zero(rs->bounce_pfn, q->bounce_pfn);
541
3cb40214
BR
542 rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
543}
544EXPORT_SYMBOL_GPL(dm_set_device_limits);
969429b5 545
3cb40214
BR
546int dm_get_device(struct dm_target *ti, const char *path, sector_t start,
547 sector_t len, int mode, struct dm_dev **result)
548{
549 int r = __table_get_device(ti->table, ti, path,
550 start, len, mode, result);
551
552 if (!r)
553 dm_set_device_limits(ti, (*result)->bdev);
1da177e4
LT
554
555 return r;
556}
557
558/*
559 * Decrement a devices use count and remove it if necessary.
560 */
561void dm_put_device(struct dm_target *ti, struct dm_dev *dd)
562{
563 if (atomic_dec_and_test(&dd->count)) {
f165921d 564 close_dev(dd, ti->table->md);
1da177e4
LT
565 list_del(&dd->list);
566 kfree(dd);
567 }
568}
569
570/*
571 * Checks to see if the target joins onto the end of the table.
572 */
573static int adjoin(struct dm_table *table, struct dm_target *ti)
574{
575 struct dm_target *prev;
576
577 if (!table->num_targets)
578 return !ti->begin;
579
580 prev = &table->targets[table->num_targets - 1];
581 return (ti->begin == (prev->begin + prev->len));
582}
583
584/*
585 * Used to dynamically allocate the arg array.
586 */
587static char **realloc_argv(unsigned *array_size, char **old_argv)
588{
589 char **argv;
590 unsigned new_size;
591
592 new_size = *array_size ? *array_size * 2 : 64;
593 argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
594 if (argv) {
595 memcpy(argv, old_argv, *array_size * sizeof(*argv));
596 *array_size = new_size;
597 }
598
599 kfree(old_argv);
600 return argv;
601}
602
603/*
604 * Destructively splits up the argument list to pass to ctr.
605 */
606int dm_split_args(int *argc, char ***argvp, char *input)
607{
608 char *start, *end = input, *out, **argv = NULL;
609 unsigned array_size = 0;
610
611 *argc = 0;
814d6862
DT
612
613 if (!input) {
614 *argvp = NULL;
615 return 0;
616 }
617
1da177e4
LT
618 argv = realloc_argv(&array_size, argv);
619 if (!argv)
620 return -ENOMEM;
621
622 while (1) {
623 start = end;
624
625 /* Skip whitespace */
626 while (*start && isspace(*start))
627 start++;
628
629 if (!*start)
630 break; /* success, we hit the end */
631
632 /* 'out' is used to remove any back-quotes */
633 end = out = start;
634 while (*end) {
635 /* Everything apart from '\0' can be quoted */
636 if (*end == '\\' && *(end + 1)) {
637 *out++ = *(end + 1);
638 end += 2;
639 continue;
640 }
641
642 if (isspace(*end))
643 break; /* end of token */
644
645 *out++ = *end++;
646 }
647
648 /* have we already filled the array ? */
649 if ((*argc + 1) > array_size) {
650 argv = realloc_argv(&array_size, argv);
651 if (!argv)
652 return -ENOMEM;
653 }
654
655 /* we know this is whitespace */
656 if (*end)
657 end++;
658
659 /* terminate the string and put it in the array */
660 *out = '\0';
661 argv[*argc] = start;
662 (*argc)++;
663 }
664
665 *argvp = argv;
666 return 0;
667}
668
669static void check_for_valid_limits(struct io_restrictions *rs)
670{
671 if (!rs->max_sectors)
defd94b7 672 rs->max_sectors = SAFE_MAX_SECTORS;
91212507
NB
673 if (!rs->max_hw_sectors)
674 rs->max_hw_sectors = SAFE_MAX_SECTORS;
1da177e4
LT
675 if (!rs->max_phys_segments)
676 rs->max_phys_segments = MAX_PHYS_SEGMENTS;
677 if (!rs->max_hw_segments)
678 rs->max_hw_segments = MAX_HW_SEGMENTS;
679 if (!rs->hardsect_size)
680 rs->hardsect_size = 1 << SECTOR_SHIFT;
681 if (!rs->max_segment_size)
682 rs->max_segment_size = MAX_SEGMENT_SIZE;
683 if (!rs->seg_boundary_mask)
684 rs->seg_boundary_mask = -1;
5ec140e6
VA
685 if (!rs->bounce_pfn)
686 rs->bounce_pfn = -1;
1da177e4
LT
687}
688
689int dm_table_add_target(struct dm_table *t, const char *type,
690 sector_t start, sector_t len, char *params)
691{
692 int r = -EINVAL, argc;
693 char **argv;
694 struct dm_target *tgt;
695
696 if ((r = check_space(t)))
697 return r;
698
699 tgt = t->targets + t->num_targets;
700 memset(tgt, 0, sizeof(*tgt));
701
702 if (!len) {
72d94861 703 DMERR("%s: zero-length target", dm_device_name(t->md));
1da177e4
LT
704 return -EINVAL;
705 }
706
707 tgt->type = dm_get_target_type(type);
708 if (!tgt->type) {
72d94861
AK
709 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
710 type);
1da177e4
LT
711 return -EINVAL;
712 }
713
714 tgt->table = t;
715 tgt->begin = start;
716 tgt->len = len;
717 tgt->error = "Unknown error";
718
719 /*
720 * Does this target adjoin the previous one ?
721 */
722 if (!adjoin(t, tgt)) {
723 tgt->error = "Gap in table";
724 r = -EINVAL;
725 goto bad;
726 }
727
728 r = dm_split_args(&argc, &argv, params);
729 if (r) {
730 tgt->error = "couldn't split parameters (insufficient memory)";
731 goto bad;
732 }
733
734 r = tgt->type->ctr(tgt, argc, argv);
735 kfree(argv);
736 if (r)
737 goto bad;
738
739 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
740
741 /* FIXME: the plan is to combine high here and then have
742 * the merge fn apply the target level restrictions. */
743 combine_restrictions_low(&t->limits, &tgt->limits);
744 return 0;
745
746 bad:
72d94861 747 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
1da177e4
LT
748 dm_put_target_type(tgt->type);
749 return r;
750}
751
752static int setup_indexes(struct dm_table *t)
753{
754 int i;
755 unsigned int total = 0;
756 sector_t *indexes;
757
758 /* allocate the space for *all* the indexes */
759 for (i = t->depth - 2; i >= 0; i--) {
760 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
761 total += t->counts[i];
762 }
763
764 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
765 if (!indexes)
766 return -ENOMEM;
767
768 /* set up internal nodes, bottom-up */
82d601dc 769 for (i = t->depth - 2; i >= 0; i--) {
1da177e4
LT
770 t->index[i] = indexes;
771 indexes += (KEYS_PER_NODE * t->counts[i]);
772 setup_btree_index(i, t);
773 }
774
775 return 0;
776}
777
778/*
779 * Builds the btree to index the map.
780 */
781int dm_table_complete(struct dm_table *t)
782{
783 int r = 0;
784 unsigned int leaf_nodes;
785
786 check_for_valid_limits(&t->limits);
787
788 /* how many indexes will the btree have ? */
789 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
790 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
791
792 /* leaf layer has already been set up */
793 t->counts[t->depth - 1] = leaf_nodes;
794 t->index[t->depth - 1] = t->highs;
795
796 if (t->depth >= 2)
797 r = setup_indexes(t);
798
799 return r;
800}
801
48c9c27b 802static DEFINE_MUTEX(_event_lock);
1da177e4
LT
803void dm_table_event_callback(struct dm_table *t,
804 void (*fn)(void *), void *context)
805{
48c9c27b 806 mutex_lock(&_event_lock);
1da177e4
LT
807 t->event_fn = fn;
808 t->event_context = context;
48c9c27b 809 mutex_unlock(&_event_lock);
1da177e4
LT
810}
811
812void dm_table_event(struct dm_table *t)
813{
814 /*
815 * You can no longer call dm_table_event() from interrupt
816 * context, use a bottom half instead.
817 */
818 BUG_ON(in_interrupt());
819
48c9c27b 820 mutex_lock(&_event_lock);
1da177e4
LT
821 if (t->event_fn)
822 t->event_fn(t->event_context);
48c9c27b 823 mutex_unlock(&_event_lock);
1da177e4
LT
824}
825
826sector_t dm_table_get_size(struct dm_table *t)
827{
828 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
829}
830
831struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
832{
14353539 833 if (index >= t->num_targets)
1da177e4
LT
834 return NULL;
835
836 return t->targets + index;
837}
838
839/*
840 * Search the btree for the correct target.
512875bd
JN
841 *
842 * Caller should check returned pointer with dm_target_is_valid()
843 * to trap I/O beyond end of device.
1da177e4
LT
844 */
845struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
846{
847 unsigned int l, n = 0, k = 0;
848 sector_t *node;
849
850 for (l = 0; l < t->depth; l++) {
851 n = get_child(n, k);
852 node = get_node(t, l, n);
853
854 for (k = 0; k < KEYS_PER_NODE; k++)
855 if (node[k] >= sector)
856 break;
857 }
858
859 return &t->targets[(KEYS_PER_NODE * n) + k];
860}
861
862void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q)
863{
864 /*
865 * Make sure we obey the optimistic sub devices
866 * restrictions.
867 */
868 blk_queue_max_sectors(q, t->limits.max_sectors);
869 q->max_phys_segments = t->limits.max_phys_segments;
870 q->max_hw_segments = t->limits.max_hw_segments;
871 q->hardsect_size = t->limits.hardsect_size;
872 q->max_segment_size = t->limits.max_segment_size;
91212507 873 q->max_hw_sectors = t->limits.max_hw_sectors;
1da177e4 874 q->seg_boundary_mask = t->limits.seg_boundary_mask;
5ec140e6 875 q->bounce_pfn = t->limits.bounce_pfn;
c9a3f6d6 876
969429b5 877 if (t->limits.no_cluster)
c9a3f6d6 878 queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
969429b5 879 else
c9a3f6d6 880 queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q);
969429b5 881
1da177e4
LT
882}
883
884unsigned int dm_table_get_num_targets(struct dm_table *t)
885{
886 return t->num_targets;
887}
888
889struct list_head *dm_table_get_devices(struct dm_table *t)
890{
891 return &t->devices;
892}
893
894int dm_table_get_mode(struct dm_table *t)
895{
896 return t->mode;
897}
898
899static void suspend_targets(struct dm_table *t, unsigned postsuspend)
900{
901 int i = t->num_targets;
902 struct dm_target *ti = t->targets;
903
904 while (i--) {
905 if (postsuspend) {
906 if (ti->type->postsuspend)
907 ti->type->postsuspend(ti);
908 } else if (ti->type->presuspend)
909 ti->type->presuspend(ti);
910
911 ti++;
912 }
913}
914
915void dm_table_presuspend_targets(struct dm_table *t)
916{
cf222b37
AK
917 if (!t)
918 return;
919
e8488d08 920 suspend_targets(t, 0);
1da177e4
LT
921}
922
923void dm_table_postsuspend_targets(struct dm_table *t)
924{
cf222b37
AK
925 if (!t)
926 return;
927
e8488d08 928 suspend_targets(t, 1);
1da177e4
LT
929}
930
8757b776 931int dm_table_resume_targets(struct dm_table *t)
1da177e4 932{
8757b776
MB
933 int i, r = 0;
934
935 for (i = 0; i < t->num_targets; i++) {
936 struct dm_target *ti = t->targets + i;
937
938 if (!ti->type->preresume)
939 continue;
940
941 r = ti->type->preresume(ti);
942 if (r)
943 return r;
944 }
1da177e4
LT
945
946 for (i = 0; i < t->num_targets; i++) {
947 struct dm_target *ti = t->targets + i;
948
949 if (ti->type->resume)
950 ti->type->resume(ti);
951 }
8757b776
MB
952
953 return 0;
1da177e4
LT
954}
955
956int dm_table_any_congested(struct dm_table *t, int bdi_bits)
957{
afb24528
PJ
958 struct dm_dev *dd;
959 struct list_head *devices = dm_table_get_devices(t);
1da177e4
LT
960 int r = 0;
961
afb24528 962 list_for_each_entry(dd, devices, list) {
165125e1 963 struct request_queue *q = bdev_get_queue(dd->bdev);
1da177e4
LT
964 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
965 }
966
967 return r;
968}
969
970void dm_table_unplug_all(struct dm_table *t)
971{
afb24528
PJ
972 struct dm_dev *dd;
973 struct list_head *devices = dm_table_get_devices(t);
1da177e4 974
afb24528 975 list_for_each_entry(dd, devices, list) {
165125e1 976 struct request_queue *q = bdev_get_queue(dd->bdev);
1da177e4 977
2ad8b1ef 978 blk_unplug(q);
1da177e4
LT
979 }
980}
981
1134e5ae
MA
982struct mapped_device *dm_table_get_md(struct dm_table *t)
983{
984 dm_get(t->md);
985
986 return t->md;
987}
988
1da177e4
LT
989EXPORT_SYMBOL(dm_vcalloc);
990EXPORT_SYMBOL(dm_get_device);
991EXPORT_SYMBOL(dm_put_device);
992EXPORT_SYMBOL(dm_table_event);
d5e404c1 993EXPORT_SYMBOL(dm_table_get_size);
1da177e4 994EXPORT_SYMBOL(dm_table_get_mode);
1134e5ae 995EXPORT_SYMBOL(dm_table_get_md);
1da177e4
LT
996EXPORT_SYMBOL(dm_table_put);
997EXPORT_SYMBOL(dm_table_get);
998EXPORT_SYMBOL(dm_table_unplug_all);