blob: 04e5fd742c2c29e42fbba920e1303b889229ccf7 [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 */
Al Viroaeb5d722008-09-02 15:28:45 -040046 fmode_t mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
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
Neil Brown91212502007-12-13 14:16:04 +0000102 lhs->max_hw_sectors =
103 min_not_zero(lhs->max_hw_sectors, rhs->max_hw_sectors);
104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 lhs->seg_boundary_mask =
106 min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask);
NeilBrown969429b2006-03-27 01:17:49 -0800107
Vasily Averin5ec140e2007-10-31 08:33:24 +0100108 lhs->bounce_pfn = min_not_zero(lhs->bounce_pfn, rhs->bounce_pfn);
109
NeilBrown969429b2006-03-27 01:17:49 -0800110 lhs->no_cluster |= rhs->no_cluster;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111}
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.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000195 * Append an empty entry to catch sectors beyond the end of
196 * the device.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 */
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000198 n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 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
Al Viroaeb5d722008-09-02 15:28:45 -0400220int dm_table_create(struct dm_table **result, fmode_t mode,
Mike Anderson1134e5a2006-03-27 01:17:54 -0800221 unsigned num_targets, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
Dmitry Monakhov094262d2007-10-19 22:38:51 +0100223 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
225 if (!t)
226 return -ENOMEM;
227
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 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;
Mike Anderson1134e5a2006-03-27 01:17:54 -0800243 t->md = md;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 *result = t;
245 return 0;
246}
247
248static void free_devices(struct list_head *devices)
249{
250 struct list_head *tmp, *next;
251
Paul Jimenezafb24522008-02-08 02:09:59 +0000252 list_for_each_safe(tmp, next, devices) {
Mikulas Patocka82b15192008-10-10 13:37:09 +0100253 struct dm_dev_internal *dd =
254 list_entry(tmp, struct dm_dev_internal, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 kfree(dd);
256 }
257}
258
Alasdair G Kergon5e198d92005-05-05 16:16:09 -0700259static void table_destroy(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
261 unsigned int i;
262
263 /* free the indexes (see dm_table_complete) */
264 if (t->depth >= 2)
265 vfree(t->index[t->depth - 2]);
266
267 /* free the targets */
268 for (i = 0; i < t->num_targets; i++) {
269 struct dm_target *tgt = t->targets + i;
270
271 if (tgt->type->dtr)
272 tgt->type->dtr(tgt);
273
274 dm_put_target_type(tgt->type);
275 }
276
277 vfree(t->highs);
278
279 /* free the device list */
280 if (t->devices.next != &t->devices) {
281 DMWARN("devices still present during destroy: "
282 "dm_table_remove_device calls missing");
283
284 free_devices(&t->devices);
285 }
286
287 kfree(t);
288}
289
290void dm_table_get(struct dm_table *t)
291{
292 atomic_inc(&t->holders);
293}
294
295void dm_table_put(struct dm_table *t)
296{
297 if (!t)
298 return;
299
300 if (atomic_dec_and_test(&t->holders))
301 table_destroy(t);
302}
303
304/*
305 * Checks to see if we need to extend highs or targets.
306 */
307static inline int check_space(struct dm_table *t)
308{
309 if (t->num_targets >= t->num_allocated)
310 return alloc_targets(t, t->num_allocated * 2);
311
312 return 0;
313}
314
315/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 * See if we've already got a device in the list.
317 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100318static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100320 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
322 list_for_each_entry (dd, l, list)
Mikulas Patocka82b15192008-10-10 13:37:09 +0100323 if (dd->dm_dev.bdev->bd_dev == dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 return dd;
325
326 return NULL;
327}
328
329/*
330 * Open a device so we can use it as a map destination.
331 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100332static int open_dev(struct dm_dev_internal *d, dev_t dev,
333 struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334{
335 static char *_claim_ptr = "I belong to device-mapper";
336 struct block_device *bdev;
337
338 int r;
339
Mikulas Patocka82b15192008-10-10 13:37:09 +0100340 BUG_ON(d->dm_dev.bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
Mikulas Patocka82b15192008-10-10 13:37:09 +0100342 bdev = open_by_devnum(dev, d->dm_dev.mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 if (IS_ERR(bdev))
344 return PTR_ERR(bdev);
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800345 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 if (r)
Al Viro9a1c3542008-02-22 20:40:24 -0500347 blkdev_put(bdev, d->dm_dev.mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 else
Mikulas Patocka82b15192008-10-10 13:37:09 +0100349 d->dm_dev.bdev = bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 return r;
351}
352
353/*
354 * Close a device that we've been using.
355 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100356static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100358 if (!d->dm_dev.bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 return;
360
Mikulas Patocka82b15192008-10-10 13:37:09 +0100361 bd_release_from_disk(d->dm_dev.bdev, dm_disk(md));
Al Viro9a1c3542008-02-22 20:40:24 -0500362 blkdev_put(d->dm_dev.bdev, d->dm_dev.mode);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100363 d->dm_dev.bdev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364}
365
366/*
Mike Anderson2cd54d92007-05-09 02:32:57 -0700367 * If possible, this checks an area of a destination device is valid.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100369static int check_device_area(struct dm_dev_internal *dd, sector_t start,
370 sector_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100372 sector_t dev_size = dd->dm_dev.bdev->bd_inode->i_size >> SECTOR_SHIFT;
Mike Anderson2cd54d92007-05-09 02:32:57 -0700373
374 if (!dev_size)
375 return 1;
376
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 return ((start < dev_size) && (len <= (dev_size - start)));
378}
379
380/*
381 * This upgrades the mode on an already open dm_dev. Being
382 * careful to leave things as they were if we fail to reopen the
383 * device.
384 */
Al Viroaeb5d722008-09-02 15:28:45 -0400385static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
Mikulas Patocka82b15192008-10-10 13:37:09 +0100386 struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387{
388 int r;
Mikulas Patocka82b15192008-10-10 13:37:09 +0100389 struct dm_dev_internal dd_copy;
390 dev_t dev = dd->dm_dev.bdev->bd_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 dd_copy = *dd;
393
Mikulas Patocka82b15192008-10-10 13:37:09 +0100394 dd->dm_dev.mode |= new_mode;
395 dd->dm_dev.bdev = NULL;
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800396 r = open_dev(dd, dev, md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 if (!r)
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800398 close_dev(&dd_copy, md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 else
400 *dd = dd_copy;
401
402 return r;
403}
404
405/*
406 * Add a device to the list, or just increment the usage count if
407 * it's already present.
408 */
409static int __table_get_device(struct dm_table *t, struct dm_target *ti,
410 const char *path, sector_t start, sector_t len,
Al Viroaeb5d722008-09-02 15:28:45 -0400411 fmode_t mode, struct dm_dev **result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412{
413 int r;
Andrew Morton69a2ce72008-02-08 02:10:14 +0000414 dev_t uninitialized_var(dev);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100415 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 unsigned int major, minor;
417
Eric Sesterhenn547bc922006-03-26 18:22:50 +0200418 BUG_ON(!t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
420 if (sscanf(path, "%u:%u", &major, &minor) == 2) {
421 /* Extract the major/minor numbers */
422 dev = MKDEV(major, minor);
423 if (MAJOR(dev) != major || MINOR(dev) != minor)
424 return -EOVERFLOW;
425 } else {
426 /* convert the path to a device */
Christoph Hellwig72e82642008-08-11 00:24:08 +0200427 struct block_device *bdev = lookup_bdev(path);
428
429 if (IS_ERR(bdev))
430 return PTR_ERR(bdev);
431 dev = bdev->bd_dev;
432 bdput(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 }
434
435 dd = find_device(&t->devices, dev);
436 if (!dd) {
437 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
438 if (!dd)
439 return -ENOMEM;
440
Mikulas Patocka82b15192008-10-10 13:37:09 +0100441 dd->dm_dev.mode = mode;
442 dd->dm_dev.bdev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800444 if ((r = open_dev(dd, dev, t->md))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 kfree(dd);
446 return r;
447 }
448
Mikulas Patocka82b15192008-10-10 13:37:09 +0100449 format_dev_t(dd->dm_dev.name, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
451 atomic_set(&dd->count, 0);
452 list_add(&dd->list, &t->devices);
453
Mikulas Patocka82b15192008-10-10 13:37:09 +0100454 } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800455 r = upgrade_mode(dd, mode, t->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 if (r)
457 return r;
458 }
459 atomic_inc(&dd->count);
460
461 if (!check_device_area(dd, start, len)) {
462 DMWARN("device %s too small for target", path);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100463 dm_put_device(ti, &dd->dm_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 return -EINVAL;
465 }
466
Mikulas Patocka82b15192008-10-10 13:37:09 +0100467 *result = &dd->dm_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
469 return 0;
470}
471
Bryn Reeves3cb40212006-10-03 01:15:42 -0700472void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
473{
Jens Axboe165125e2007-07-24 09:28:11 +0200474 struct request_queue *q = bdev_get_queue(bdev);
Bryn Reeves3cb40212006-10-03 01:15:42 -0700475 struct io_restrictions *rs = &ti->limits;
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100476 char b[BDEVNAME_SIZE];
477
478 if (unlikely(!q)) {
479 DMWARN("%s: Cannot set limits for nonexistent device %s",
480 dm_device_name(ti->table->md), bdevname(bdev, b));
481 return;
482 }
Bryn Reeves3cb40212006-10-03 01:15:42 -0700483
484 /*
485 * Combine the device limits low.
486 *
487 * FIXME: if we move an io_restriction struct
488 * into q this would just be a call to
489 * combine_restrictions_low()
490 */
491 rs->max_sectors =
492 min_not_zero(rs->max_sectors, q->max_sectors);
493
Milan Broz9980c632008-07-21 12:00:39 +0100494 /*
495 * Check if merge fn is supported.
496 * If not we'll force DM to use PAGE_SIZE or
497 * smaller I/O, just to be safe.
Bryn Reeves3cb40212006-10-03 01:15:42 -0700498 */
Milan Broz9980c632008-07-21 12:00:39 +0100499
500 if (q->merge_bvec_fn && !ti->type->merge)
Bryn Reeves3cb40212006-10-03 01:15:42 -0700501 rs->max_sectors =
502 min_not_zero(rs->max_sectors,
503 (unsigned int) (PAGE_SIZE >> 9));
504
505 rs->max_phys_segments =
506 min_not_zero(rs->max_phys_segments,
507 q->max_phys_segments);
508
509 rs->max_hw_segments =
510 min_not_zero(rs->max_hw_segments, q->max_hw_segments);
511
512 rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size);
513
514 rs->max_segment_size =
515 min_not_zero(rs->max_segment_size, q->max_segment_size);
516
Neil Brown91212502007-12-13 14:16:04 +0000517 rs->max_hw_sectors =
518 min_not_zero(rs->max_hw_sectors, q->max_hw_sectors);
519
Bryn Reeves3cb40212006-10-03 01:15:42 -0700520 rs->seg_boundary_mask =
521 min_not_zero(rs->seg_boundary_mask,
522 q->seg_boundary_mask);
523
Vasily Averin5ec140e2007-10-31 08:33:24 +0100524 rs->bounce_pfn = min_not_zero(rs->bounce_pfn, q->bounce_pfn);
525
Bryn Reeves3cb40212006-10-03 01:15:42 -0700526 rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
527}
528EXPORT_SYMBOL_GPL(dm_set_device_limits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
530int dm_get_device(struct dm_target *ti, const char *path, sector_t start,
Al Viroaeb5d722008-09-02 15:28:45 -0400531 sector_t len, fmode_t mode, struct dm_dev **result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
533 int r = __table_get_device(ti->table, ti, path,
534 start, len, mode, result);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535
Bryn Reeves3cb40212006-10-03 01:15:42 -0700536 if (!r)
537 dm_set_device_limits(ti, (*result)->bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
539 return r;
540}
541
542/*
543 * Decrement a devices use count and remove it if necessary.
544 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100545void dm_put_device(struct dm_target *ti, struct dm_dev *d)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100547 struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
548 dm_dev);
549
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 if (atomic_dec_and_test(&dd->count)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800551 close_dev(dd, ti->table->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 list_del(&dd->list);
553 kfree(dd);
554 }
555}
556
557/*
558 * Checks to see if the target joins onto the end of the table.
559 */
560static int adjoin(struct dm_table *table, struct dm_target *ti)
561{
562 struct dm_target *prev;
563
564 if (!table->num_targets)
565 return !ti->begin;
566
567 prev = &table->targets[table->num_targets - 1];
568 return (ti->begin == (prev->begin + prev->len));
569}
570
571/*
572 * Used to dynamically allocate the arg array.
573 */
574static char **realloc_argv(unsigned *array_size, char **old_argv)
575{
576 char **argv;
577 unsigned new_size;
578
579 new_size = *array_size ? *array_size * 2 : 64;
580 argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
581 if (argv) {
582 memcpy(argv, old_argv, *array_size * sizeof(*argv));
583 *array_size = new_size;
584 }
585
586 kfree(old_argv);
587 return argv;
588}
589
590/*
591 * Destructively splits up the argument list to pass to ctr.
592 */
593int dm_split_args(int *argc, char ***argvp, char *input)
594{
595 char *start, *end = input, *out, **argv = NULL;
596 unsigned array_size = 0;
597
598 *argc = 0;
David Teigland814d6862006-06-26 00:27:31 -0700599
600 if (!input) {
601 *argvp = NULL;
602 return 0;
603 }
604
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 argv = realloc_argv(&array_size, argv);
606 if (!argv)
607 return -ENOMEM;
608
609 while (1) {
610 start = end;
611
612 /* Skip whitespace */
613 while (*start && isspace(*start))
614 start++;
615
616 if (!*start)
617 break; /* success, we hit the end */
618
619 /* 'out' is used to remove any back-quotes */
620 end = out = start;
621 while (*end) {
622 /* Everything apart from '\0' can be quoted */
623 if (*end == '\\' && *(end + 1)) {
624 *out++ = *(end + 1);
625 end += 2;
626 continue;
627 }
628
629 if (isspace(*end))
630 break; /* end of token */
631
632 *out++ = *end++;
633 }
634
635 /* have we already filled the array ? */
636 if ((*argc + 1) > array_size) {
637 argv = realloc_argv(&array_size, argv);
638 if (!argv)
639 return -ENOMEM;
640 }
641
642 /* we know this is whitespace */
643 if (*end)
644 end++;
645
646 /* terminate the string and put it in the array */
647 *out = '\0';
648 argv[*argc] = start;
649 (*argc)++;
650 }
651
652 *argvp = argv;
653 return 0;
654}
655
656static void check_for_valid_limits(struct io_restrictions *rs)
657{
658 if (!rs->max_sectors)
Mike Christiedefd94b2005-12-05 02:37:06 -0600659 rs->max_sectors = SAFE_MAX_SECTORS;
Neil Brown91212502007-12-13 14:16:04 +0000660 if (!rs->max_hw_sectors)
661 rs->max_hw_sectors = SAFE_MAX_SECTORS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 if (!rs->max_phys_segments)
663 rs->max_phys_segments = MAX_PHYS_SEGMENTS;
664 if (!rs->max_hw_segments)
665 rs->max_hw_segments = MAX_HW_SEGMENTS;
666 if (!rs->hardsect_size)
667 rs->hardsect_size = 1 << SECTOR_SHIFT;
668 if (!rs->max_segment_size)
669 rs->max_segment_size = MAX_SEGMENT_SIZE;
670 if (!rs->seg_boundary_mask)
Milan Broz0e435ac2008-12-03 12:55:08 +0100671 rs->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
Vasily Averin5ec140e2007-10-31 08:33:24 +0100672 if (!rs->bounce_pfn)
673 rs->bounce_pfn = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674}
675
676int dm_table_add_target(struct dm_table *t, const char *type,
677 sector_t start, sector_t len, char *params)
678{
679 int r = -EINVAL, argc;
680 char **argv;
681 struct dm_target *tgt;
682
683 if ((r = check_space(t)))
684 return r;
685
686 tgt = t->targets + t->num_targets;
687 memset(tgt, 0, sizeof(*tgt));
688
689 if (!len) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700690 DMERR("%s: zero-length target", dm_device_name(t->md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 return -EINVAL;
692 }
693
694 tgt->type = dm_get_target_type(type);
695 if (!tgt->type) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700696 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
697 type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 return -EINVAL;
699 }
700
701 tgt->table = t;
702 tgt->begin = start;
703 tgt->len = len;
704 tgt->error = "Unknown error";
705
706 /*
707 * Does this target adjoin the previous one ?
708 */
709 if (!adjoin(t, tgt)) {
710 tgt->error = "Gap in table";
711 r = -EINVAL;
712 goto bad;
713 }
714
715 r = dm_split_args(&argc, &argv, params);
716 if (r) {
717 tgt->error = "couldn't split parameters (insufficient memory)";
718 goto bad;
719 }
720
721 r = tgt->type->ctr(tgt, argc, argv);
722 kfree(argv);
723 if (r)
724 goto bad;
725
726 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
727
728 /* FIXME: the plan is to combine high here and then have
729 * the merge fn apply the target level restrictions. */
730 combine_restrictions_low(&t->limits, &tgt->limits);
731 return 0;
732
733 bad:
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700734 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 dm_put_target_type(tgt->type);
736 return r;
737}
738
739static int setup_indexes(struct dm_table *t)
740{
741 int i;
742 unsigned int total = 0;
743 sector_t *indexes;
744
745 /* allocate the space for *all* the indexes */
746 for (i = t->depth - 2; i >= 0; i--) {
747 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
748 total += t->counts[i];
749 }
750
751 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
752 if (!indexes)
753 return -ENOMEM;
754
755 /* set up internal nodes, bottom-up */
Jun'ichi Nomura82d601d2008-02-08 02:10:04 +0000756 for (i = t->depth - 2; i >= 0; i--) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 t->index[i] = indexes;
758 indexes += (KEYS_PER_NODE * t->counts[i]);
759 setup_btree_index(i, t);
760 }
761
762 return 0;
763}
764
765/*
766 * Builds the btree to index the map.
767 */
768int dm_table_complete(struct dm_table *t)
769{
770 int r = 0;
771 unsigned int leaf_nodes;
772
773 check_for_valid_limits(&t->limits);
774
775 /* how many indexes will the btree have ? */
776 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
777 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
778
779 /* leaf layer has already been set up */
780 t->counts[t->depth - 1] = leaf_nodes;
781 t->index[t->depth - 1] = t->highs;
782
783 if (t->depth >= 2)
784 r = setup_indexes(t);
785
786 return r;
787}
788
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800789static DEFINE_MUTEX(_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790void dm_table_event_callback(struct dm_table *t,
791 void (*fn)(void *), void *context)
792{
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800793 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 t->event_fn = fn;
795 t->event_context = context;
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800796 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797}
798
799void dm_table_event(struct dm_table *t)
800{
801 /*
802 * You can no longer call dm_table_event() from interrupt
803 * context, use a bottom half instead.
804 */
805 BUG_ON(in_interrupt());
806
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800807 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 if (t->event_fn)
809 t->event_fn(t->event_context);
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800810 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811}
812
813sector_t dm_table_get_size(struct dm_table *t)
814{
815 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
816}
817
818struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
819{
Milan Broz14353532006-06-26 00:27:27 -0700820 if (index >= t->num_targets)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 return NULL;
822
823 return t->targets + index;
824}
825
826/*
827 * Search the btree for the correct target.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000828 *
829 * Caller should check returned pointer with dm_target_is_valid()
830 * to trap I/O beyond end of device.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 */
832struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
833{
834 unsigned int l, n = 0, k = 0;
835 sector_t *node;
836
837 for (l = 0; l < t->depth; l++) {
838 n = get_child(n, k);
839 node = get_node(t, l, n);
840
841 for (k = 0; k < KEYS_PER_NODE; k++)
842 if (node[k] >= sector)
843 break;
844 }
845
846 return &t->targets[(KEYS_PER_NODE * n) + k];
847}
848
849void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q)
850{
851 /*
852 * Make sure we obey the optimistic sub devices
853 * restrictions.
854 */
855 blk_queue_max_sectors(q, t->limits.max_sectors);
856 q->max_phys_segments = t->limits.max_phys_segments;
857 q->max_hw_segments = t->limits.max_hw_segments;
858 q->hardsect_size = t->limits.hardsect_size;
859 q->max_segment_size = t->limits.max_segment_size;
Neil Brown91212502007-12-13 14:16:04 +0000860 q->max_hw_sectors = t->limits.max_hw_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 q->seg_boundary_mask = t->limits.seg_boundary_mask;
Vasily Averin5ec140e2007-10-31 08:33:24 +0100862 q->bounce_pfn = t->limits.bounce_pfn;
Jens Axboec9a3f6d2008-04-29 19:12:35 +0200863
NeilBrown969429b2006-03-27 01:17:49 -0800864 if (t->limits.no_cluster)
Jens Axboec9a3f6d2008-04-29 19:12:35 +0200865 queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
NeilBrown969429b2006-03-27 01:17:49 -0800866 else
Jens Axboec9a3f6d2008-04-29 19:12:35 +0200867 queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q);
NeilBrown969429b2006-03-27 01:17:49 -0800868
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869}
870
871unsigned int dm_table_get_num_targets(struct dm_table *t)
872{
873 return t->num_targets;
874}
875
876struct list_head *dm_table_get_devices(struct dm_table *t)
877{
878 return &t->devices;
879}
880
Al Viroaeb5d722008-09-02 15:28:45 -0400881fmode_t dm_table_get_mode(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
883 return t->mode;
884}
885
886static void suspend_targets(struct dm_table *t, unsigned postsuspend)
887{
888 int i = t->num_targets;
889 struct dm_target *ti = t->targets;
890
891 while (i--) {
892 if (postsuspend) {
893 if (ti->type->postsuspend)
894 ti->type->postsuspend(ti);
895 } else if (ti->type->presuspend)
896 ti->type->presuspend(ti);
897
898 ti++;
899 }
900}
901
902void dm_table_presuspend_targets(struct dm_table *t)
903{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -0700904 if (!t)
905 return;
906
Adrian Bunke8488d02008-04-24 22:10:51 +0100907 suspend_targets(t, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908}
909
910void dm_table_postsuspend_targets(struct dm_table *t)
911{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -0700912 if (!t)
913 return;
914
Adrian Bunke8488d02008-04-24 22:10:51 +0100915 suspend_targets(t, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
Milan Broz8757b772006-10-03 01:15:36 -0700918int dm_table_resume_targets(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919{
Milan Broz8757b772006-10-03 01:15:36 -0700920 int i, r = 0;
921
922 for (i = 0; i < t->num_targets; i++) {
923 struct dm_target *ti = t->targets + i;
924
925 if (!ti->type->preresume)
926 continue;
927
928 r = ti->type->preresume(ti);
929 if (r)
930 return r;
931 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932
933 for (i = 0; i < t->num_targets; i++) {
934 struct dm_target *ti = t->targets + i;
935
936 if (ti->type->resume)
937 ti->type->resume(ti);
938 }
Milan Broz8757b772006-10-03 01:15:36 -0700939
940 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941}
942
943int dm_table_any_congested(struct dm_table *t, int bdi_bits)
944{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100945 struct dm_dev_internal *dd;
Paul Jimenezafb24522008-02-08 02:09:59 +0000946 struct list_head *devices = dm_table_get_devices(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 int r = 0;
948
Paul Jimenezafb24522008-02-08 02:09:59 +0000949 list_for_each_entry(dd, devices, list) {
Mikulas Patocka82b15192008-10-10 13:37:09 +0100950 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100951 char b[BDEVNAME_SIZE];
952
953 if (likely(q))
954 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
955 else
956 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
957 dm_device_name(t->md),
958 bdevname(dd->dm_dev.bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 }
960
961 return r;
962}
963
964void dm_table_unplug_all(struct dm_table *t)
965{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100966 struct dm_dev_internal *dd;
Paul Jimenezafb24522008-02-08 02:09:59 +0000967 struct list_head *devices = dm_table_get_devices(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
Paul Jimenezafb24522008-02-08 02:09:59 +0000969 list_for_each_entry(dd, devices, list) {
Mikulas Patocka82b15192008-10-10 13:37:09 +0100970 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100971 char b[BDEVNAME_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100973 if (likely(q))
974 blk_unplug(q);
975 else
976 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
977 dm_device_name(t->md),
978 bdevname(dd->dm_dev.bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 }
980}
981
Mike Anderson1134e5a2006-03-27 01:17:54 -0800982struct mapped_device *dm_table_get_md(struct dm_table *t)
983{
984 dm_get(t->md);
985
986 return t->md;
987}
988
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989EXPORT_SYMBOL(dm_vcalloc);
990EXPORT_SYMBOL(dm_get_device);
991EXPORT_SYMBOL(dm_put_device);
992EXPORT_SYMBOL(dm_table_event);
Alasdair G Kergond5e404c2005-07-12 15:53:05 -0700993EXPORT_SYMBOL(dm_table_get_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994EXPORT_SYMBOL(dm_table_get_mode);
Mike Anderson1134e5a2006-03-27 01:17:54 -0800995EXPORT_SYMBOL(dm_table_get_md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996EXPORT_SYMBOL(dm_table_put);
997EXPORT_SYMBOL(dm_table_get);
998EXPORT_SYMBOL(dm_table_unplug_all);