blob: 9924ea23032d6f7418f15f1f8b4198b5c5dedd88 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright (C) 2001 Sistina Software (UK) Limited.
Mikulas Patockad5816872009-01-06 03:05:10 +00003 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
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>
André Goddard Rosae7d28602009-12-14 18:01:06 -080015#include <linux/string.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/slab.h>
17#include <linux/interrupt.h>
Arjan van de Ven48c9c272006-03-27 01:18:20 -080018#include <linux/mutex.h>
Mikulas Patockad5816872009-01-06 03:05:10 +000019#include <linux/delay.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <asm/atomic.h>
21
Alasdair G Kergon72d94862006-06-26 00:27:35 -070022#define DM_MSG_PREFIX "table"
23
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#define MAX_DEPTH 16
25#define NODE_SIZE L1_CACHE_BYTES
26#define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
27#define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
28
Mikulas Patockad5816872009-01-06 03:05:10 +000029/*
30 * The table has always exactly one reference from either mapped_device->map
31 * or hash_cell->new_map. This reference is not counted in table->holders.
32 * A pair of dm_create_table/dm_destroy_table functions is used for table
33 * creation/destruction.
34 *
35 * Temporary references from the other code increase table->holders. A pair
36 * of dm_table_get/dm_table_put functions is used to manipulate it.
37 *
38 * When the table is about to be destroyed, we wait for table->holders to
39 * drop to zero.
40 */
41
Linus Torvalds1da177e2005-04-16 15:20:36 -070042struct dm_table {
Mike Anderson1134e5a2006-03-27 01:17:54 -080043 struct mapped_device *md;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044 atomic_t holders;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +010045 unsigned type;
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47 /* btree table */
48 unsigned int depth;
49 unsigned int counts[MAX_DEPTH]; /* in nodes */
50 sector_t *index[MAX_DEPTH];
51
52 unsigned int num_targets;
53 unsigned int num_allocated;
54 sector_t *highs;
55 struct dm_target *targets;
56
57 /*
58 * Indicates the rw permissions for the new logical
59 * device. This should be a combination of FMODE_READ
60 * and FMODE_WRITE.
61 */
Al Viroaeb5d722008-09-02 15:28:45 -040062 fmode_t mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64 /* a list of devices used by this table */
65 struct list_head devices;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 /* events get handed up using this callback */
68 void (*event_fn)(void *);
69 void *event_context;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +010070
71 struct dm_md_mempools *mempools;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072};
73
74/*
75 * Similar to ceiling(log_size(n))
76 */
77static unsigned int int_log(unsigned int n, unsigned int base)
78{
79 int result = 0;
80
81 while (n > 1) {
82 n = dm_div_up(n, base);
83 result++;
84 }
85
86 return result;
87}
88
89/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 * Calculate the index of the child node of the n'th node k'th key.
91 */
92static inline unsigned int get_child(unsigned int n, unsigned int k)
93{
94 return (n * CHILDREN_PER_NODE) + k;
95}
96
97/*
98 * Return the n'th node of level l from table t.
99 */
100static inline sector_t *get_node(struct dm_table *t,
101 unsigned int l, unsigned int n)
102{
103 return t->index[l] + (n * KEYS_PER_NODE);
104}
105
106/*
107 * Return the highest key that you could lookup from the n'th
108 * node on level l of the btree.
109 */
110static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
111{
112 for (; l < t->depth - 1; l++)
113 n = get_child(n, CHILDREN_PER_NODE - 1);
114
115 if (n >= t->counts[l])
116 return (sector_t) - 1;
117
118 return get_node(t, l, n)[KEYS_PER_NODE - 1];
119}
120
121/*
122 * Fills in a level of the btree based on the highs of the level
123 * below it.
124 */
125static int setup_btree_index(unsigned int l, struct dm_table *t)
126{
127 unsigned int n, k;
128 sector_t *node;
129
130 for (n = 0U; n < t->counts[l]; n++) {
131 node = get_node(t, l, n);
132
133 for (k = 0U; k < KEYS_PER_NODE; k++)
134 node[k] = high(t, l + 1, get_child(n, k));
135 }
136
137 return 0;
138}
139
140void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
141{
142 unsigned long size;
143 void *addr;
144
145 /*
146 * Check that we're not going to overflow.
147 */
148 if (nmemb > (ULONG_MAX / elem_size))
149 return NULL;
150
151 size = nmemb * elem_size;
152 addr = vmalloc(size);
153 if (addr)
154 memset(addr, 0, size);
155
156 return addr;
157}
158
159/*
160 * highs, and targets are managed as dynamic arrays during a
161 * table load.
162 */
163static int alloc_targets(struct dm_table *t, unsigned int num)
164{
165 sector_t *n_highs;
166 struct dm_target *n_targets;
167 int n = t->num_targets;
168
169 /*
170 * Allocate both the target array and offset array at once.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000171 * Append an empty entry to catch sectors beyond the end of
172 * the device.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 */
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000174 n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 sizeof(sector_t));
176 if (!n_highs)
177 return -ENOMEM;
178
179 n_targets = (struct dm_target *) (n_highs + num);
180
181 if (n) {
182 memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
183 memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
184 }
185
186 memset(n_highs + n, -1, sizeof(*n_highs) * (num - n));
187 vfree(t->highs);
188
189 t->num_allocated = num;
190 t->highs = n_highs;
191 t->targets = n_targets;
192
193 return 0;
194}
195
Al Viroaeb5d722008-09-02 15:28:45 -0400196int dm_table_create(struct dm_table **result, fmode_t mode,
Mike Anderson1134e5a2006-03-27 01:17:54 -0800197 unsigned num_targets, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198{
Dmitry Monakhov094262d2007-10-19 22:38:51 +0100199 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
201 if (!t)
202 return -ENOMEM;
203
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 INIT_LIST_HEAD(&t->devices);
Mikulas Patockad5816872009-01-06 03:05:10 +0000205 atomic_set(&t->holders, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
207 if (!num_targets)
208 num_targets = KEYS_PER_NODE;
209
210 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
211
212 if (alloc_targets(t, num_targets)) {
213 kfree(t);
214 t = NULL;
215 return -ENOMEM;
216 }
217
218 t->mode = mode;
Mike Anderson1134e5a2006-03-27 01:17:54 -0800219 t->md = md;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 *result = t;
221 return 0;
222}
223
224static void free_devices(struct list_head *devices)
225{
226 struct list_head *tmp, *next;
227
Paul Jimenezafb24522008-02-08 02:09:59 +0000228 list_for_each_safe(tmp, next, devices) {
Mikulas Patocka82b15192008-10-10 13:37:09 +0100229 struct dm_dev_internal *dd =
230 list_entry(tmp, struct dm_dev_internal, list);
Jonthan Brassow1b6da752009-06-22 10:12:29 +0100231 DMWARN("dm_table_destroy: dm_put_device call missing for %s",
232 dd->dm_dev.name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 kfree(dd);
234 }
235}
236
Mikulas Patockad5816872009-01-06 03:05:10 +0000237void dm_table_destroy(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238{
239 unsigned int i;
240
Alasdair G Kergona7940152009-12-10 23:52:23 +0000241 if (!t)
242 return;
243
Mikulas Patockad5816872009-01-06 03:05:10 +0000244 while (atomic_read(&t->holders))
245 msleep(1);
246 smp_mb();
247
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 /* free the indexes (see dm_table_complete) */
249 if (t->depth >= 2)
250 vfree(t->index[t->depth - 2]);
251
252 /* free the targets */
253 for (i = 0; i < t->num_targets; i++) {
254 struct dm_target *tgt = t->targets + i;
255
256 if (tgt->type->dtr)
257 tgt->type->dtr(tgt);
258
259 dm_put_target_type(tgt->type);
260 }
261
262 vfree(t->highs);
263
264 /* free the device list */
Jonthan Brassow1b6da752009-06-22 10:12:29 +0100265 if (t->devices.next != &t->devices)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 free_devices(&t->devices);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100268 dm_free_md_mempools(t->mempools);
269
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 kfree(t);
271}
272
273void dm_table_get(struct dm_table *t)
274{
275 atomic_inc(&t->holders);
276}
277
278void dm_table_put(struct dm_table *t)
279{
280 if (!t)
281 return;
282
Mikulas Patockad5816872009-01-06 03:05:10 +0000283 smp_mb__before_atomic_dec();
284 atomic_dec(&t->holders);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285}
286
287/*
288 * Checks to see if we need to extend highs or targets.
289 */
290static inline int check_space(struct dm_table *t)
291{
292 if (t->num_targets >= t->num_allocated)
293 return alloc_targets(t, t->num_allocated * 2);
294
295 return 0;
296}
297
298/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * See if we've already got a device in the list.
300 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100301static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100303 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
305 list_for_each_entry (dd, l, list)
Mikulas Patocka82b15192008-10-10 13:37:09 +0100306 if (dd->dm_dev.bdev->bd_dev == dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 return dd;
308
309 return NULL;
310}
311
312/*
313 * Open a device so we can use it as a map destination.
314 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100315static int open_dev(struct dm_dev_internal *d, dev_t dev,
316 struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317{
318 static char *_claim_ptr = "I belong to device-mapper";
319 struct block_device *bdev;
320
321 int r;
322
Mikulas Patocka82b15192008-10-10 13:37:09 +0100323 BUG_ON(d->dm_dev.bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
Mikulas Patocka82b15192008-10-10 13:37:09 +0100325 bdev = open_by_devnum(dev, d->dm_dev.mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 if (IS_ERR(bdev))
327 return PTR_ERR(bdev);
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800328 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 if (r)
Al Viro9a1c3542008-02-22 20:40:24 -0500330 blkdev_put(bdev, d->dm_dev.mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 else
Mikulas Patocka82b15192008-10-10 13:37:09 +0100332 d->dm_dev.bdev = bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 return r;
334}
335
336/*
337 * Close a device that we've been using.
338 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100339static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100341 if (!d->dm_dev.bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 return;
343
Mikulas Patocka82b15192008-10-10 13:37:09 +0100344 bd_release_from_disk(d->dm_dev.bdev, dm_disk(md));
Al Viro9a1c3542008-02-22 20:40:24 -0500345 blkdev_put(d->dm_dev.bdev, d->dm_dev.mode);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100346 d->dm_dev.bdev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347}
348
349/*
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100350 * If possible, this checks an area of a destination device is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 */
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100352static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
353 sector_t start, sector_t len, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100355 struct queue_limits *limits = data;
356 struct block_device *bdev = dev->bdev;
357 sector_t dev_size =
358 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100359 unsigned short logical_block_size_sectors =
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100360 limits->logical_block_size >> SECTOR_SHIFT;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100361 char b[BDEVNAME_SIZE];
Mike Anderson2cd54d92007-05-09 02:32:57 -0700362
363 if (!dev_size)
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100364 return 0;
Mike Anderson2cd54d92007-05-09 02:32:57 -0700365
Mike Snitzer5dea2712009-07-23 20:30:42 +0100366 if ((start >= dev_size) || (start + len > dev_size)) {
Mike Snitzera963a952009-09-04 20:40:24 +0100367 DMWARN("%s: %s too small for target: "
368 "start=%llu, len=%llu, dev_size=%llu",
369 dm_device_name(ti->table->md), bdevname(bdev, b),
370 (unsigned long long)start,
371 (unsigned long long)len,
372 (unsigned long long)dev_size);
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100373 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100374 }
375
376 if (logical_block_size_sectors <= 1)
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100377 return 0;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100378
379 if (start & (logical_block_size_sectors - 1)) {
380 DMWARN("%s: start=%llu not aligned to h/w "
Mike Snitzera963a952009-09-04 20:40:24 +0100381 "logical block size %u of %s",
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100382 dm_device_name(ti->table->md),
383 (unsigned long long)start,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100384 limits->logical_block_size, bdevname(bdev, b));
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100385 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100386 }
387
Mike Snitzer5dea2712009-07-23 20:30:42 +0100388 if (len & (logical_block_size_sectors - 1)) {
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100389 DMWARN("%s: len=%llu not aligned to h/w "
Mike Snitzera963a952009-09-04 20:40:24 +0100390 "logical block size %u of %s",
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100391 dm_device_name(ti->table->md),
Mike Snitzer5dea2712009-07-23 20:30:42 +0100392 (unsigned long long)len,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100393 limits->logical_block_size, bdevname(bdev, b));
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100394 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100395 }
396
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100397 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398}
399
400/*
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100401 * This upgrades the mode on an already open dm_dev, being
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 * careful to leave things as they were if we fail to reopen the
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100403 * device and not to touch the existing bdev field in case
404 * it is accessed concurrently inside dm_table_any_congested().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 */
Al Viroaeb5d722008-09-02 15:28:45 -0400406static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
Mikulas Patocka82b15192008-10-10 13:37:09 +0100407 struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408{
409 int r;
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100410 struct dm_dev_internal dd_new, dd_old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100412 dd_new = dd_old = *dd;
413
414 dd_new.dm_dev.mode |= new_mode;
415 dd_new.dm_dev.bdev = NULL;
416
417 r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md);
418 if (r)
419 return r;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Mikulas Patocka82b15192008-10-10 13:37:09 +0100421 dd->dm_dev.mode |= new_mode;
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100422 close_dev(&dd_old, md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100424 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425}
426
427/*
428 * Add a device to the list, or just increment the usage count if
429 * it's already present.
430 */
431static int __table_get_device(struct dm_table *t, struct dm_target *ti,
Nikanth Karthikesan8215d6e2010-03-06 02:32:27 +0000432 const char *path, fmode_t mode, struct dm_dev **result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
434 int r;
Andrew Morton69a2ce72008-02-08 02:10:14 +0000435 dev_t uninitialized_var(dev);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100436 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 unsigned int major, minor;
438
Eric Sesterhenn547bc922006-03-26 18:22:50 +0200439 BUG_ON(!t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
441 if (sscanf(path, "%u:%u", &major, &minor) == 2) {
442 /* Extract the major/minor numbers */
443 dev = MKDEV(major, minor);
444 if (MAJOR(dev) != major || MINOR(dev) != minor)
445 return -EOVERFLOW;
446 } else {
447 /* convert the path to a device */
Christoph Hellwig72e82642008-08-11 00:24:08 +0200448 struct block_device *bdev = lookup_bdev(path);
449
450 if (IS_ERR(bdev))
451 return PTR_ERR(bdev);
452 dev = bdev->bd_dev;
453 bdput(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 }
455
456 dd = find_device(&t->devices, dev);
457 if (!dd) {
458 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
459 if (!dd)
460 return -ENOMEM;
461
Mikulas Patocka82b15192008-10-10 13:37:09 +0100462 dd->dm_dev.mode = mode;
463 dd->dm_dev.bdev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800465 if ((r = open_dev(dd, dev, t->md))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 kfree(dd);
467 return r;
468 }
469
Mikulas Patocka82b15192008-10-10 13:37:09 +0100470 format_dev_t(dd->dm_dev.name, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 atomic_set(&dd->count, 0);
473 list_add(&dd->list, &t->devices);
474
Mikulas Patocka82b15192008-10-10 13:37:09 +0100475 } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800476 r = upgrade_mode(dd, mode, t->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 if (r)
478 return r;
479 }
480 atomic_inc(&dd->count);
481
Mikulas Patocka82b15192008-10-10 13:37:09 +0100482 *result = &dd->dm_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 return 0;
484}
485
Mike Snitzer5ab97582009-06-22 10:12:32 +0100486/*
487 * Returns the minimum that is _not_ zero, unless both are zero.
488 */
489#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))
490
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100491int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
Mike Snitzer5dea2712009-07-23 20:30:42 +0100492 sector_t start, sector_t len, void *data)
Bryn Reeves3cb40212006-10-03 01:15:42 -0700493{
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100494 struct queue_limits *limits = data;
495 struct block_device *bdev = dev->bdev;
Jens Axboe165125e2007-07-24 09:28:11 +0200496 struct request_queue *q = bdev_get_queue(bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100497 char b[BDEVNAME_SIZE];
498
499 if (unlikely(!q)) {
500 DMWARN("%s: Cannot set limits for nonexistent device %s",
501 dm_device_name(ti->table->md), bdevname(bdev, b));
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100502 return 0;
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100503 }
Bryn Reeves3cb40212006-10-03 01:15:42 -0700504
Martin K. Petersenb27d7f12010-01-11 03:21:50 -0500505 if (bdev_stack_limits(limits, bdev, start) < 0)
506 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
Mike Snitzera963a952009-09-04 20:40:24 +0100507 "physical_block_size=%u, logical_block_size=%u, "
508 "alignment_offset=%u, start=%llu",
509 dm_device_name(ti->table->md), bdevname(bdev, b),
510 q->limits.physical_block_size,
511 q->limits.logical_block_size,
512 q->limits.alignment_offset,
Martin K. Petersenb27d7f12010-01-11 03:21:50 -0500513 (unsigned long long) start << SECTOR_SHIFT);
Bryn Reeves3cb40212006-10-03 01:15:42 -0700514
Milan Broz9980c632008-07-21 12:00:39 +0100515 /*
516 * Check if merge fn is supported.
517 * If not we'll force DM to use PAGE_SIZE or
518 * smaller I/O, just to be safe.
Bryn Reeves3cb40212006-10-03 01:15:42 -0700519 */
Milan Broz9980c632008-07-21 12:00:39 +0100520
521 if (q->merge_bvec_fn && !ti->type->merge)
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100522 limits->max_sectors =
523 min_not_zero(limits->max_sectors,
Bryn Reeves3cb40212006-10-03 01:15:42 -0700524 (unsigned int) (PAGE_SIZE >> 9));
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100525 return 0;
Bryn Reeves3cb40212006-10-03 01:15:42 -0700526}
527EXPORT_SYMBOL_GPL(dm_set_device_limits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
Nikanth Karthikesan8215d6e2010-03-06 02:32:27 +0000529int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
530 struct dm_dev **result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
Nikanth Karthikesan8215d6e2010-03-06 02:32:27 +0000532 return __table_get_device(ti->table, ti, path, mode, result);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533}
534
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536/*
537 * Decrement a devices use count and remove it if necessary.
538 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100539void dm_put_device(struct dm_target *ti, struct dm_dev *d)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100541 struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
542 dm_dev);
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 if (atomic_dec_and_test(&dd->count)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800545 close_dev(dd, ti->table->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 list_del(&dd->list);
547 kfree(dd);
548 }
549}
550
551/*
552 * Checks to see if the target joins onto the end of the table.
553 */
554static int adjoin(struct dm_table *table, struct dm_target *ti)
555{
556 struct dm_target *prev;
557
558 if (!table->num_targets)
559 return !ti->begin;
560
561 prev = &table->targets[table->num_targets - 1];
562 return (ti->begin == (prev->begin + prev->len));
563}
564
565/*
566 * Used to dynamically allocate the arg array.
567 */
568static char **realloc_argv(unsigned *array_size, char **old_argv)
569{
570 char **argv;
571 unsigned new_size;
572
573 new_size = *array_size ? *array_size * 2 : 64;
574 argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
575 if (argv) {
576 memcpy(argv, old_argv, *array_size * sizeof(*argv));
577 *array_size = new_size;
578 }
579
580 kfree(old_argv);
581 return argv;
582}
583
584/*
585 * Destructively splits up the argument list to pass to ctr.
586 */
587int dm_split_args(int *argc, char ***argvp, char *input)
588{
589 char *start, *end = input, *out, **argv = NULL;
590 unsigned array_size = 0;
591
592 *argc = 0;
David Teigland814d6862006-06-26 00:27:31 -0700593
594 if (!input) {
595 *argvp = NULL;
596 return 0;
597 }
598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 argv = realloc_argv(&array_size, argv);
600 if (!argv)
601 return -ENOMEM;
602
603 while (1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 /* Skip whitespace */
André Goddard Rosae7d28602009-12-14 18:01:06 -0800605 start = skip_spaces(end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
607 if (!*start)
608 break; /* success, we hit the end */
609
610 /* 'out' is used to remove any back-quotes */
611 end = out = start;
612 while (*end) {
613 /* Everything apart from '\0' can be quoted */
614 if (*end == '\\' && *(end + 1)) {
615 *out++ = *(end + 1);
616 end += 2;
617 continue;
618 }
619
620 if (isspace(*end))
621 break; /* end of token */
622
623 *out++ = *end++;
624 }
625
626 /* have we already filled the array ? */
627 if ((*argc + 1) > array_size) {
628 argv = realloc_argv(&array_size, argv);
629 if (!argv)
630 return -ENOMEM;
631 }
632
633 /* we know this is whitespace */
634 if (*end)
635 end++;
636
637 /* terminate the string and put it in the array */
638 *out = '\0';
639 argv[*argc] = start;
640 (*argc)++;
641 }
642
643 *argvp = argv;
644 return 0;
645}
646
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100647/*
648 * Impose necessary and sufficient conditions on a devices's table such
649 * that any incoming bio which respects its logical_block_size can be
650 * processed successfully. If it falls across the boundary between
651 * two or more targets, the size of each piece it gets split into must
652 * be compatible with the logical_block_size of the target processing it.
653 */
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100654static int validate_hardware_logical_block_alignment(struct dm_table *table,
655 struct queue_limits *limits)
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100656{
657 /*
658 * This function uses arithmetic modulo the logical_block_size
659 * (in units of 512-byte sectors).
660 */
661 unsigned short device_logical_block_size_sects =
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100662 limits->logical_block_size >> SECTOR_SHIFT;
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100663
664 /*
665 * Offset of the start of the next table entry, mod logical_block_size.
666 */
667 unsigned short next_target_start = 0;
668
669 /*
670 * Given an aligned bio that extends beyond the end of a
671 * target, how many sectors must the next target handle?
672 */
673 unsigned short remaining = 0;
674
675 struct dm_target *uninitialized_var(ti);
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100676 struct queue_limits ti_limits;
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100677 unsigned i = 0;
678
679 /*
680 * Check each entry in the table in turn.
681 */
682 while (i < dm_table_get_num_targets(table)) {
683 ti = dm_table_get_target(table, i++);
684
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100685 blk_set_default_limits(&ti_limits);
686
687 /* combine all target devices' limits */
688 if (ti->type->iterate_devices)
689 ti->type->iterate_devices(ti, dm_set_device_limits,
690 &ti_limits);
691
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100692 /*
693 * If the remaining sectors fall entirely within this
694 * table entry are they compatible with its logical_block_size?
695 */
696 if (remaining < ti->len &&
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100697 remaining & ((ti_limits.logical_block_size >>
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100698 SECTOR_SHIFT) - 1))
699 break; /* Error */
700
701 next_target_start =
702 (unsigned short) ((next_target_start + ti->len) &
703 (device_logical_block_size_sects - 1));
704 remaining = next_target_start ?
705 device_logical_block_size_sects - next_target_start : 0;
706 }
707
708 if (remaining) {
709 DMWARN("%s: table line %u (start sect %llu len %llu) "
Mike Snitzera963a952009-09-04 20:40:24 +0100710 "not aligned to h/w logical block size %u",
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100711 dm_device_name(table->md), i,
712 (unsigned long long) ti->begin,
713 (unsigned long long) ti->len,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100714 limits->logical_block_size);
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100715 return -EINVAL;
716 }
717
718 return 0;
719}
720
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721int dm_table_add_target(struct dm_table *t, const char *type,
722 sector_t start, sector_t len, char *params)
723{
724 int r = -EINVAL, argc;
725 char **argv;
726 struct dm_target *tgt;
727
728 if ((r = check_space(t)))
729 return r;
730
731 tgt = t->targets + t->num_targets;
732 memset(tgt, 0, sizeof(*tgt));
733
734 if (!len) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700735 DMERR("%s: zero-length target", dm_device_name(t->md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 return -EINVAL;
737 }
738
739 tgt->type = dm_get_target_type(type);
740 if (!tgt->type) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700741 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
742 type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 return -EINVAL;
744 }
745
746 tgt->table = t;
747 tgt->begin = start;
748 tgt->len = len;
749 tgt->error = "Unknown error";
750
751 /*
752 * Does this target adjoin the previous one ?
753 */
754 if (!adjoin(t, tgt)) {
755 tgt->error = "Gap in table";
756 r = -EINVAL;
757 goto bad;
758 }
759
760 r = dm_split_args(&argc, &argv, params);
761 if (r) {
762 tgt->error = "couldn't split parameters (insufficient memory)";
763 goto bad;
764 }
765
766 r = tgt->type->ctr(tgt, argc, argv);
767 kfree(argv);
768 if (r)
769 goto bad;
770
771 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
772
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 return 0;
774
775 bad:
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700776 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 dm_put_target_type(tgt->type);
778 return r;
779}
780
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100781int dm_table_set_type(struct dm_table *t)
782{
783 unsigned i;
784 unsigned bio_based = 0, request_based = 0;
785 struct dm_target *tgt;
786 struct dm_dev_internal *dd;
787 struct list_head *devices;
788
789 for (i = 0; i < t->num_targets; i++) {
790 tgt = t->targets + i;
791 if (dm_target_request_based(tgt))
792 request_based = 1;
793 else
794 bio_based = 1;
795
796 if (bio_based && request_based) {
797 DMWARN("Inconsistent table: different target types"
798 " can't be mixed up");
799 return -EINVAL;
800 }
801 }
802
803 if (bio_based) {
804 /* We must use this table as bio-based */
805 t->type = DM_TYPE_BIO_BASED;
806 return 0;
807 }
808
809 BUG_ON(!request_based); /* No targets in this table */
810
811 /* Non-request-stackable devices can't be used for request-based dm */
812 devices = dm_table_get_devices(t);
813 list_for_each_entry(dd, devices, list) {
814 if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) {
815 DMWARN("table load rejected: including"
816 " non-request-stackable devices");
817 return -EINVAL;
818 }
819 }
820
821 /*
822 * Request-based dm supports only tables that have a single target now.
823 * To support multiple targets, request splitting support is needed,
824 * and that needs lots of changes in the block-layer.
825 * (e.g. request completion process for partial completion.)
826 */
827 if (t->num_targets > 1) {
828 DMWARN("Request-based dm doesn't support multiple targets yet");
829 return -EINVAL;
830 }
831
832 t->type = DM_TYPE_REQUEST_BASED;
833
834 return 0;
835}
836
837unsigned dm_table_get_type(struct dm_table *t)
838{
839 return t->type;
840}
841
842bool dm_table_request_based(struct dm_table *t)
843{
844 return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED;
845}
846
847int dm_table_alloc_md_mempools(struct dm_table *t)
848{
849 unsigned type = dm_table_get_type(t);
850
851 if (unlikely(type == DM_TYPE_NONE)) {
852 DMWARN("no table type is set, can't allocate mempools");
853 return -EINVAL;
854 }
855
856 t->mempools = dm_alloc_md_mempools(type);
857 if (!t->mempools)
858 return -ENOMEM;
859
860 return 0;
861}
862
863void dm_table_free_md_mempools(struct dm_table *t)
864{
865 dm_free_md_mempools(t->mempools);
866 t->mempools = NULL;
867}
868
869struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
870{
871 return t->mempools;
872}
873
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874static int setup_indexes(struct dm_table *t)
875{
876 int i;
877 unsigned int total = 0;
878 sector_t *indexes;
879
880 /* allocate the space for *all* the indexes */
881 for (i = t->depth - 2; i >= 0; i--) {
882 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
883 total += t->counts[i];
884 }
885
886 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
887 if (!indexes)
888 return -ENOMEM;
889
890 /* set up internal nodes, bottom-up */
Jun'ichi Nomura82d601d2008-02-08 02:10:04 +0000891 for (i = t->depth - 2; i >= 0; i--) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 t->index[i] = indexes;
893 indexes += (KEYS_PER_NODE * t->counts[i]);
894 setup_btree_index(i, t);
895 }
896
897 return 0;
898}
899
900/*
901 * Builds the btree to index the map.
902 */
903int dm_table_complete(struct dm_table *t)
904{
905 int r = 0;
906 unsigned int leaf_nodes;
907
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 /* how many indexes will the btree have ? */
909 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
910 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
911
912 /* leaf layer has already been set up */
913 t->counts[t->depth - 1] = leaf_nodes;
914 t->index[t->depth - 1] = t->highs;
915
916 if (t->depth >= 2)
917 r = setup_indexes(t);
918
919 return r;
920}
921
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800922static DEFINE_MUTEX(_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923void dm_table_event_callback(struct dm_table *t,
924 void (*fn)(void *), void *context)
925{
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800926 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 t->event_fn = fn;
928 t->event_context = context;
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800929 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930}
931
932void dm_table_event(struct dm_table *t)
933{
934 /*
935 * You can no longer call dm_table_event() from interrupt
936 * context, use a bottom half instead.
937 */
938 BUG_ON(in_interrupt());
939
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800940 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 if (t->event_fn)
942 t->event_fn(t->event_context);
Arjan van de Ven48c9c272006-03-27 01:18:20 -0800943 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944}
945
946sector_t dm_table_get_size(struct dm_table *t)
947{
948 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
949}
950
951struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
952{
Milan Broz14353532006-06-26 00:27:27 -0700953 if (index >= t->num_targets)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 return NULL;
955
956 return t->targets + index;
957}
958
959/*
960 * Search the btree for the correct target.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000961 *
962 * Caller should check returned pointer with dm_target_is_valid()
963 * to trap I/O beyond end of device.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 */
965struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
966{
967 unsigned int l, n = 0, k = 0;
968 sector_t *node;
969
970 for (l = 0; l < t->depth; l++) {
971 n = get_child(n, k);
972 node = get_node(t, l, n);
973
974 for (k = 0; k < KEYS_PER_NODE; k++)
975 if (node[k] >= sector)
976 break;
977 }
978
979 return &t->targets[(KEYS_PER_NODE * n) + k];
980}
981
Martin K. Petersen9c470082009-04-09 00:27:12 +0100982/*
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100983 * Establish the new table's queue_limits and validate them.
984 */
985int dm_calculate_queue_limits(struct dm_table *table,
986 struct queue_limits *limits)
987{
988 struct dm_target *uninitialized_var(ti);
989 struct queue_limits ti_limits;
990 unsigned i = 0;
991
992 blk_set_default_limits(limits);
993
994 while (i < dm_table_get_num_targets(table)) {
995 blk_set_default_limits(&ti_limits);
996
997 ti = dm_table_get_target(table, i++);
998
999 if (!ti->type->iterate_devices)
1000 goto combine_limits;
1001
1002 /*
1003 * Combine queue limits of all the devices this target uses.
1004 */
1005 ti->type->iterate_devices(ti, dm_set_device_limits,
1006 &ti_limits);
1007
Mike Snitzer40bea432009-09-04 20:40:25 +01001008 /* Set I/O hints portion of queue limits */
1009 if (ti->type->io_hints)
1010 ti->type->io_hints(ti, &ti_limits);
1011
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001012 /*
1013 * Check each device area is consistent with the target's
1014 * overall queue limits.
1015 */
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +01001016 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1017 &ti_limits))
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001018 return -EINVAL;
1019
1020combine_limits:
1021 /*
1022 * Merge this target's queue limits into the overall limits
1023 * for the table.
1024 */
1025 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
Martin K. Petersenb27d7f12010-01-11 03:21:50 -05001026 DMWARN("%s: adding target device "
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001027 "(start sect %llu len %llu) "
Martin K. Petersenb27d7f12010-01-11 03:21:50 -05001028 "caused an alignment inconsistency",
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001029 dm_device_name(table->md),
1030 (unsigned long long) ti->begin,
1031 (unsigned long long) ti->len);
1032 }
1033
1034 return validate_hardware_logical_block_alignment(table, limits);
1035}
1036
1037/*
Martin K. Petersen9c470082009-04-09 00:27:12 +01001038 * Set the integrity profile for this device if all devices used have
1039 * matching profiles.
1040 */
1041static void dm_table_set_integrity(struct dm_table *t)
1042{
1043 struct list_head *devices = dm_table_get_devices(t);
1044 struct dm_dev_internal *prev = NULL, *dd = NULL;
1045
1046 if (!blk_get_integrity(dm_disk(t->md)))
1047 return;
1048
1049 list_for_each_entry(dd, devices, list) {
1050 if (prev &&
1051 blk_integrity_compare(prev->dm_dev.bdev->bd_disk,
1052 dd->dm_dev.bdev->bd_disk) < 0) {
1053 DMWARN("%s: integrity not set: %s and %s mismatch",
1054 dm_device_name(t->md),
1055 prev->dm_dev.bdev->bd_disk->disk_name,
1056 dd->dm_dev.bdev->bd_disk->disk_name);
1057 goto no_integrity;
1058 }
1059 prev = dd;
1060 }
1061
1062 if (!prev || !bdev_get_integrity(prev->dm_dev.bdev))
1063 goto no_integrity;
1064
1065 blk_integrity_register(dm_disk(t->md),
1066 bdev_get_integrity(prev->dm_dev.bdev));
1067
1068 return;
1069
1070no_integrity:
1071 blk_integrity_register(dm_disk(t->md), NULL);
1072
1073 return;
1074}
1075
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001076void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1077 struct queue_limits *limits)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078{
1079 /*
Mike Snitzer11977642009-06-22 10:12:32 +01001080 * Copy table's limits to the DM device's request_queue
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 */
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001082 q->limits = *limits;
Jens Axboec9a3f6d2008-04-29 19:12:35 +02001083
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001084 if (limits->no_cluster)
Jens Axboec9a3f6d2008-04-29 19:12:35 +02001085 queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
NeilBrown969429b2006-03-27 01:17:49 -08001086 else
Jens Axboec9a3f6d2008-04-29 19:12:35 +02001087 queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q);
NeilBrown969429b2006-03-27 01:17:49 -08001088
Martin K. Petersen9c470082009-04-09 00:27:12 +01001089 dm_table_set_integrity(t);
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +01001090
1091 /*
1092 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1093 * visible to other CPUs because, once the flag is set, incoming bios
1094 * are processed by request-based dm, which refers to the queue
1095 * settings.
1096 * Until the flag set, bios are passed to bio-based dm and queued to
1097 * md->deferred where queue settings are not needed yet.
1098 * Those bios are passed to request-based dm at the resume time.
1099 */
1100 smp_mb();
1101 if (dm_table_request_based(t))
1102 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103}
1104
1105unsigned int dm_table_get_num_targets(struct dm_table *t)
1106{
1107 return t->num_targets;
1108}
1109
1110struct list_head *dm_table_get_devices(struct dm_table *t)
1111{
1112 return &t->devices;
1113}
1114
Al Viroaeb5d722008-09-02 15:28:45 -04001115fmode_t dm_table_get_mode(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116{
1117 return t->mode;
1118}
1119
1120static void suspend_targets(struct dm_table *t, unsigned postsuspend)
1121{
1122 int i = t->num_targets;
1123 struct dm_target *ti = t->targets;
1124
1125 while (i--) {
1126 if (postsuspend) {
1127 if (ti->type->postsuspend)
1128 ti->type->postsuspend(ti);
1129 } else if (ti->type->presuspend)
1130 ti->type->presuspend(ti);
1131
1132 ti++;
1133 }
1134}
1135
1136void dm_table_presuspend_targets(struct dm_table *t)
1137{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -07001138 if (!t)
1139 return;
1140
Adrian Bunke8488d02008-04-24 22:10:51 +01001141 suspend_targets(t, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142}
1143
1144void dm_table_postsuspend_targets(struct dm_table *t)
1145{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -07001146 if (!t)
1147 return;
1148
Adrian Bunke8488d02008-04-24 22:10:51 +01001149 suspend_targets(t, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150}
1151
Milan Broz8757b772006-10-03 01:15:36 -07001152int dm_table_resume_targets(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153{
Milan Broz8757b772006-10-03 01:15:36 -07001154 int i, r = 0;
1155
1156 for (i = 0; i < t->num_targets; i++) {
1157 struct dm_target *ti = t->targets + i;
1158
1159 if (!ti->type->preresume)
1160 continue;
1161
1162 r = ti->type->preresume(ti);
1163 if (r)
1164 return r;
1165 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
1167 for (i = 0; i < t->num_targets; i++) {
1168 struct dm_target *ti = t->targets + i;
1169
1170 if (ti->type->resume)
1171 ti->type->resume(ti);
1172 }
Milan Broz8757b772006-10-03 01:15:36 -07001173
1174 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175}
1176
1177int dm_table_any_congested(struct dm_table *t, int bdi_bits)
1178{
Mikulas Patocka82b15192008-10-10 13:37:09 +01001179 struct dm_dev_internal *dd;
Paul Jimenezafb24522008-02-08 02:09:59 +00001180 struct list_head *devices = dm_table_get_devices(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 int r = 0;
1182
Paul Jimenezafb24522008-02-08 02:09:59 +00001183 list_for_each_entry(dd, devices, list) {
Mikulas Patocka82b15192008-10-10 13:37:09 +01001184 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +01001185 char b[BDEVNAME_SIZE];
1186
1187 if (likely(q))
1188 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
1189 else
1190 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1191 dm_device_name(t->md),
1192 bdevname(dd->dm_dev.bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 }
1194
1195 return r;
1196}
1197
Kiyoshi Uedacec47e32009-06-22 10:12:35 +01001198int dm_table_any_busy_target(struct dm_table *t)
1199{
1200 unsigned i;
1201 struct dm_target *ti;
1202
1203 for (i = 0; i < t->num_targets; i++) {
1204 ti = t->targets + i;
1205 if (ti->type->busy && ti->type->busy(ti))
1206 return 1;
1207 }
1208
1209 return 0;
1210}
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212void dm_table_unplug_all(struct dm_table *t)
1213{
Mikulas Patocka82b15192008-10-10 13:37:09 +01001214 struct dm_dev_internal *dd;
Paul Jimenezafb24522008-02-08 02:09:59 +00001215 struct list_head *devices = dm_table_get_devices(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216
Paul Jimenezafb24522008-02-08 02:09:59 +00001217 list_for_each_entry(dd, devices, list) {
Mikulas Patocka82b15192008-10-10 13:37:09 +01001218 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +01001219 char b[BDEVNAME_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +01001221 if (likely(q))
1222 blk_unplug(q);
1223 else
1224 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
1225 dm_device_name(t->md),
1226 bdevname(dd->dm_dev.bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 }
1228}
1229
Mike Anderson1134e5a2006-03-27 01:17:54 -08001230struct mapped_device *dm_table_get_md(struct dm_table *t)
1231{
Mike Anderson1134e5a2006-03-27 01:17:54 -08001232 return t->md;
1233}
1234
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235EXPORT_SYMBOL(dm_vcalloc);
1236EXPORT_SYMBOL(dm_get_device);
1237EXPORT_SYMBOL(dm_put_device);
1238EXPORT_SYMBOL(dm_table_event);
Alasdair G Kergond5e404c2005-07-12 15:53:05 -07001239EXPORT_SYMBOL(dm_table_get_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240EXPORT_SYMBOL(dm_table_get_mode);
Mike Anderson1134e5a2006-03-27 01:17:54 -08001241EXPORT_SYMBOL(dm_table_get_md);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242EXPORT_SYMBOL(dm_table_put);
1243EXPORT_SYMBOL(dm_table_get);
1244EXPORT_SYMBOL(dm_table_unplug_all);