blob: b2bd1ebf4562902aba2b747c742a5f4da3f659e1 [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>
Arun Sharma600634972011-07-26 16:09:06 -070020#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
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
29struct dm_table {
Mike Anderson1134e5a2006-03-27 01:17:54 -080030 struct mapped_device *md;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +010031 unsigned type;
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
33 /* btree table */
34 unsigned int depth;
35 unsigned int counts[MAX_DEPTH]; /* in nodes */
36 sector_t *index[MAX_DEPTH];
37
38 unsigned int num_targets;
39 unsigned int num_allocated;
40 sector_t *highs;
41 struct dm_target *targets;
42
Alasdair G Kergon36a04562011-10-31 20:19:04 +000043 struct target_type *immutable_target_type;
Martin K. Petersena91a2782011-03-17 11:11:05 +010044 unsigned integrity_supported:1;
Alasdair G Kergon3791e2f2011-10-31 20:19:00 +000045 unsigned singleton:1;
Mike Snitzer5ae89a82010-08-12 04:14:08 +010046
Linus Torvalds1da177e2005-04-16 15:20:36 -070047 /*
48 * Indicates the rw permissions for the new logical
49 * device. This should be a combination of FMODE_READ
50 * and FMODE_WRITE.
51 */
Al Viroaeb5d722008-09-02 15:28:45 -040052 fmode_t mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
54 /* a list of devices used by this table */
55 struct list_head devices;
56
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 /* events get handed up using this callback */
58 void (*event_fn)(void *);
59 void *event_context;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +010060
61 struct dm_md_mempools *mempools;
NeilBrown9d357b02011-01-13 20:00:01 +000062
63 struct list_head target_callbacks;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064};
65
66/*
67 * Similar to ceiling(log_size(n))
68 */
69static unsigned int int_log(unsigned int n, unsigned int base)
70{
71 int result = 0;
72
73 while (n > 1) {
74 n = dm_div_up(n, base);
75 result++;
76 }
77
78 return result;
79}
80
81/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 * Calculate the index of the child node of the n'th node k'th key.
83 */
84static inline unsigned int get_child(unsigned int n, unsigned int k)
85{
86 return (n * CHILDREN_PER_NODE) + k;
87}
88
89/*
90 * Return the n'th node of level l from table t.
91 */
92static inline sector_t *get_node(struct dm_table *t,
93 unsigned int l, unsigned int n)
94{
95 return t->index[l] + (n * KEYS_PER_NODE);
96}
97
98/*
99 * Return the highest key that you could lookup from the n'th
100 * node on level l of the btree.
101 */
102static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
103{
104 for (; l < t->depth - 1; l++)
105 n = get_child(n, CHILDREN_PER_NODE - 1);
106
107 if (n >= t->counts[l])
108 return (sector_t) - 1;
109
110 return get_node(t, l, n)[KEYS_PER_NODE - 1];
111}
112
113/*
114 * Fills in a level of the btree based on the highs of the level
115 * below it.
116 */
117static int setup_btree_index(unsigned int l, struct dm_table *t)
118{
119 unsigned int n, k;
120 sector_t *node;
121
122 for (n = 0U; n < t->counts[l]; n++) {
123 node = get_node(t, l, n);
124
125 for (k = 0U; k < KEYS_PER_NODE; k++)
126 node[k] = high(t, l + 1, get_child(n, k));
127 }
128
129 return 0;
130}
131
132void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
133{
134 unsigned long size;
135 void *addr;
136
137 /*
138 * Check that we're not going to overflow.
139 */
140 if (nmemb > (ULONG_MAX / elem_size))
141 return NULL;
142
143 size = nmemb * elem_size;
Joe Perchese29e65a2011-08-02 12:32:02 +0100144 addr = vzalloc(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
146 return addr;
147}
Mike Snitzer08649012011-08-02 12:32:04 +0100148EXPORT_SYMBOL(dm_vcalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
150/*
151 * highs, and targets are managed as dynamic arrays during a
152 * table load.
153 */
154static int alloc_targets(struct dm_table *t, unsigned int num)
155{
156 sector_t *n_highs;
157 struct dm_target *n_targets;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159 /*
160 * Allocate both the target array and offset array at once.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000161 * Append an empty entry to catch sectors beyond the end of
162 * the device.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 */
Jun'ichi Nomura512875b2007-12-13 14:15:25 +0000164 n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 sizeof(sector_t));
166 if (!n_highs)
167 return -ENOMEM;
168
169 n_targets = (struct dm_target *) (n_highs + num);
170
Mikulas Patocka57a2f232013-11-22 19:51:39 -0500171 memset(n_highs, -1, sizeof(*n_highs) * num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 vfree(t->highs);
173
174 t->num_allocated = num;
175 t->highs = n_highs;
176 t->targets = n_targets;
177
178 return 0;
179}
180
Al Viroaeb5d722008-09-02 15:28:45 -0400181int dm_table_create(struct dm_table **result, fmode_t mode,
Mike Anderson1134e5a2006-03-27 01:17:54 -0800182 unsigned num_targets, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183{
Dmitry Monakhov094262d2007-10-19 22:38:51 +0100184 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
186 if (!t)
187 return -ENOMEM;
188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 INIT_LIST_HEAD(&t->devices);
NeilBrown9d357b02011-01-13 20:00:01 +0000190 INIT_LIST_HEAD(&t->target_callbacks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191
192 if (!num_targets)
193 num_targets = KEYS_PER_NODE;
194
195 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
196
Mikulas Patocka5b2d0652013-11-22 19:52:06 -0500197 if (!num_targets) {
198 kfree(t);
199 return -ENOMEM;
200 }
201
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 if (alloc_targets(t, num_targets)) {
203 kfree(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 return -ENOMEM;
205 }
206
207 t->mode = mode;
Mike Anderson1134e5a2006-03-27 01:17:54 -0800208 t->md = md;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 *result = t;
210 return 0;
211}
212
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500213static void free_devices(struct list_head *devices, struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 struct list_head *tmp, *next;
216
Paul Jimenezafb24522008-02-08 02:09:59 +0000217 list_for_each_safe(tmp, next, devices) {
Mikulas Patocka82b15192008-10-10 13:37:09 +0100218 struct dm_dev_internal *dd =
219 list_entry(tmp, struct dm_dev_internal, list);
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500220 DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s",
221 dm_device_name(md), dd->dm_dev->name);
222 dm_put_table_device(md, dd->dm_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 kfree(dd);
224 }
225}
226
Mikulas Patockad5816872009-01-06 03:05:10 +0000227void dm_table_destroy(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228{
229 unsigned int i;
230
Alasdair G Kergona7940152009-12-10 23:52:23 +0000231 if (!t)
232 return;
233
Will Drewry26803b92010-08-12 04:14:03 +0100234 /* free the indexes */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 if (t->depth >= 2)
236 vfree(t->index[t->depth - 2]);
237
238 /* free the targets */
239 for (i = 0; i < t->num_targets; i++) {
240 struct dm_target *tgt = t->targets + i;
241
242 if (tgt->type->dtr)
243 tgt->type->dtr(tgt);
244
245 dm_put_target_type(tgt->type);
246 }
247
248 vfree(t->highs);
249
250 /* free the device list */
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500251 free_devices(&t->devices, t->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100253 dm_free_md_mempools(t->mempools);
254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 kfree(t);
256}
257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 * See if we've already got a device in the list.
260 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100261static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262{
Mikulas Patocka82b15192008-10-10 13:37:09 +0100263 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
265 list_for_each_entry (dd, l, list)
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500266 if (dd->dm_dev->bdev->bd_dev == dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 return dd;
268
269 return NULL;
270}
271
272/*
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100273 * If possible, this checks an area of a destination device is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 */
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100275static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
276 sector_t start, sector_t len, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277{
Milan Brozf4808ca2011-05-29 13:02:52 +0100278 struct request_queue *q;
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100279 struct queue_limits *limits = data;
280 struct block_device *bdev = dev->bdev;
281 sector_t dev_size =
282 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100283 unsigned short logical_block_size_sectors =
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100284 limits->logical_block_size >> SECTOR_SHIFT;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100285 char b[BDEVNAME_SIZE];
Mike Anderson2cd54d92007-05-09 02:32:57 -0700286
Milan Brozf4808ca2011-05-29 13:02:52 +0100287 /*
288 * Some devices exist without request functions,
289 * such as loop devices not yet bound to backing files.
290 * Forbid the use of such devices.
291 */
292 q = bdev_get_queue(bdev);
293 if (!q || !q->make_request_fn) {
294 DMWARN("%s: %s is not yet initialised: "
295 "start=%llu, len=%llu, dev_size=%llu",
296 dm_device_name(ti->table->md), bdevname(bdev, b),
297 (unsigned long long)start,
298 (unsigned long long)len,
299 (unsigned long long)dev_size);
300 return 1;
301 }
302
Mike Anderson2cd54d92007-05-09 02:32:57 -0700303 if (!dev_size)
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100304 return 0;
Mike Anderson2cd54d92007-05-09 02:32:57 -0700305
Mike Snitzer5dea2712009-07-23 20:30:42 +0100306 if ((start >= dev_size) || (start + len > dev_size)) {
Mike Snitzera963a952009-09-04 20:40:24 +0100307 DMWARN("%s: %s too small for target: "
308 "start=%llu, len=%llu, dev_size=%llu",
309 dm_device_name(ti->table->md), bdevname(bdev, b),
310 (unsigned long long)start,
311 (unsigned long long)len,
312 (unsigned long long)dev_size);
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100313 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100314 }
315
316 if (logical_block_size_sectors <= 1)
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100317 return 0;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100318
319 if (start & (logical_block_size_sectors - 1)) {
320 DMWARN("%s: start=%llu not aligned to h/w "
Mike Snitzera963a952009-09-04 20:40:24 +0100321 "logical block size %u of %s",
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100322 dm_device_name(ti->table->md),
323 (unsigned long long)start,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100324 limits->logical_block_size, bdevname(bdev, b));
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100325 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100326 }
327
Mike Snitzer5dea2712009-07-23 20:30:42 +0100328 if (len & (logical_block_size_sectors - 1)) {
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100329 DMWARN("%s: len=%llu not aligned to h/w "
Mike Snitzera963a952009-09-04 20:40:24 +0100330 "logical block size %u of %s",
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100331 dm_device_name(ti->table->md),
Mike Snitzer5dea2712009-07-23 20:30:42 +0100332 (unsigned long long)len,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100333 limits->logical_block_size, bdevname(bdev, b));
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100334 return 1;
Mike Snitzer02acc3a2009-06-22 10:12:30 +0100335 }
336
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +0100337 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338}
339
340/*
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100341 * This upgrades the mode on an already open dm_dev, being
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 * careful to leave things as they were if we fail to reopen the
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100343 * device and not to touch the existing bdev field in case
344 * it is accessed concurrently inside dm_table_any_congested().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 */
Al Viroaeb5d722008-09-02 15:28:45 -0400346static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
Mikulas Patocka82b15192008-10-10 13:37:09 +0100347 struct mapped_device *md)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
349 int r;
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500350 struct dm_dev *old_dev, *new_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500352 old_dev = dd->dm_dev;
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100353
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500354 r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
355 dd->dm_dev->mode | new_mode, &new_dev);
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100356 if (r)
357 return r;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500359 dd->dm_dev = new_dev;
360 dm_put_table_device(md, old_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Alasdair G Kergon570b9d92009-04-02 19:55:28 +0100362 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363}
364
365/*
366 * Add a device to the list, or just increment the usage count if
367 * it's already present.
368 */
Mike Snitzer08649012011-08-02 12:32:04 +0100369int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
370 struct dm_dev **result)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
372 int r;
Andrew Morton69a2ce72008-02-08 02:10:14 +0000373 dev_t uninitialized_var(dev);
Mikulas Patocka82b15192008-10-10 13:37:09 +0100374 struct dm_dev_internal *dd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 unsigned int major, minor;
Mike Snitzer08649012011-08-02 12:32:04 +0100376 struct dm_table *t = ti->table;
Mikulas Patocka31998ef2012-03-28 18:41:26 +0100377 char dummy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
Eric Sesterhenn547bc922006-03-26 18:22:50 +0200379 BUG_ON(!t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
Mikulas Patocka31998ef2012-03-28 18:41:26 +0100381 if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 /* Extract the major/minor numbers */
383 dev = MKDEV(major, minor);
384 if (MAJOR(dev) != major || MINOR(dev) != minor)
385 return -EOVERFLOW;
386 } else {
387 /* convert the path to a device */
Christoph Hellwig72e82642008-08-11 00:24:08 +0200388 struct block_device *bdev = lookup_bdev(path);
389
390 if (IS_ERR(bdev))
391 return PTR_ERR(bdev);
392 dev = bdev->bd_dev;
393 bdput(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 }
395
396 dd = find_device(&t->devices, dev);
397 if (!dd) {
398 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
399 if (!dd)
400 return -ENOMEM;
401
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500402 if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 kfree(dd);
404 return r;
405 }
406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 atomic_set(&dd->count, 0);
408 list_add(&dd->list, &t->devices);
409
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500410 } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
Jun'ichi Nomuraf1659212006-03-27 01:17:59 -0800411 r = upgrade_mode(dd, mode, t->md);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 if (r)
413 return r;
414 }
415 atomic_inc(&dd->count);
416
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500417 *result = dd->dm_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 return 0;
419}
Mike Snitzer08649012011-08-02 12:32:04 +0100420EXPORT_SYMBOL(dm_get_device);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
Mike Snitzer11f04312014-06-03 10:30:28 -0400422static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
423 sector_t start, sector_t len, void *data)
Bryn Reeves3cb40212006-10-03 01:15:42 -0700424{
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100425 struct queue_limits *limits = data;
426 struct block_device *bdev = dev->bdev;
Jens Axboe165125e2007-07-24 09:28:11 +0200427 struct request_queue *q = bdev_get_queue(bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100428 char b[BDEVNAME_SIZE];
429
430 if (unlikely(!q)) {
431 DMWARN("%s: Cannot set limits for nonexistent device %s",
432 dm_device_name(ti->table->md), bdevname(bdev, b));
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100433 return 0;
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +0100434 }
Bryn Reeves3cb40212006-10-03 01:15:42 -0700435
Martin K. Petersenb27d7f12010-01-11 03:21:50 -0500436 if (bdev_stack_limits(limits, bdev, start) < 0)
437 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
Mike Snitzera963a952009-09-04 20:40:24 +0100438 "physical_block_size=%u, logical_block_size=%u, "
439 "alignment_offset=%u, start=%llu",
440 dm_device_name(ti->table->md), bdevname(bdev, b),
441 q->limits.physical_block_size,
442 q->limits.logical_block_size,
443 q->limits.alignment_offset,
Martin K. Petersenb27d7f12010-01-11 03:21:50 -0500444 (unsigned long long) start << SECTOR_SHIFT);
Bryn Reeves3cb40212006-10-03 01:15:42 -0700445
Milan Broz9980c632008-07-21 12:00:39 +0100446 /*
447 * Check if merge fn is supported.
448 * If not we'll force DM to use PAGE_SIZE or
449 * smaller I/O, just to be safe.
Bryn Reeves3cb40212006-10-03 01:15:42 -0700450 */
Mikulas Patockad5b9dd02011-08-02 12:32:04 +0100451 if (dm_queue_merge_is_compulsory(q) && !ti->type->merge)
Mike Snitzer72d4cd92010-12-17 08:34:20 +0100452 blk_limits_max_hw_sectors(limits,
453 (unsigned int) (PAGE_SIZE >> 9));
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100454 return 0;
Bryn Reeves3cb40212006-10-03 01:15:42 -0700455}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457/*
Mike Snitzer08649012011-08-02 12:32:04 +0100458 * Decrement a device's use count and remove it if necessary.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 */
Mikulas Patocka82b15192008-10-10 13:37:09 +0100460void dm_put_device(struct dm_target *ti, struct dm_dev *d)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461{
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500462 int found = 0;
463 struct list_head *devices = &ti->table->devices;
464 struct dm_dev_internal *dd;
Mikulas Patocka82b15192008-10-10 13:37:09 +0100465
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500466 list_for_each_entry(dd, devices, list) {
467 if (dd->dm_dev == d) {
468 found = 1;
469 break;
470 }
471 }
472 if (!found) {
473 DMWARN("%s: device %s not in table devices list",
474 dm_device_name(ti->table->md), d->name);
475 return;
476 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 if (atomic_dec_and_test(&dd->count)) {
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500478 dm_put_table_device(ti->table->md, d);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 list_del(&dd->list);
480 kfree(dd);
481 }
482}
Mike Snitzer08649012011-08-02 12:32:04 +0100483EXPORT_SYMBOL(dm_put_device);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485/*
486 * Checks to see if the target joins onto the end of the table.
487 */
488static int adjoin(struct dm_table *table, struct dm_target *ti)
489{
490 struct dm_target *prev;
491
492 if (!table->num_targets)
493 return !ti->begin;
494
495 prev = &table->targets[table->num_targets - 1];
496 return (ti->begin == (prev->begin + prev->len));
497}
498
499/*
500 * Used to dynamically allocate the arg array.
Mikulas Patockaf36afb32013-10-31 13:55:45 -0400501 *
502 * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
503 * process messages even if some device is suspended. These messages have a
504 * small fixed number of arguments.
505 *
506 * On the other hand, dm-switch needs to process bulk data using messages and
507 * excessive use of GFP_NOIO could cause trouble.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 */
509static char **realloc_argv(unsigned *array_size, char **old_argv)
510{
511 char **argv;
512 unsigned new_size;
Mikulas Patockaf36afb32013-10-31 13:55:45 -0400513 gfp_t gfp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Mikulas Patockaf36afb32013-10-31 13:55:45 -0400515 if (*array_size) {
516 new_size = *array_size * 2;
517 gfp = GFP_KERNEL;
518 } else {
519 new_size = 8;
520 gfp = GFP_NOIO;
521 }
522 argv = kmalloc(new_size * sizeof(*argv), gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 if (argv) {
524 memcpy(argv, old_argv, *array_size * sizeof(*argv));
525 *array_size = new_size;
526 }
527
528 kfree(old_argv);
529 return argv;
530}
531
532/*
533 * Destructively splits up the argument list to pass to ctr.
534 */
535int dm_split_args(int *argc, char ***argvp, char *input)
536{
537 char *start, *end = input, *out, **argv = NULL;
538 unsigned array_size = 0;
539
540 *argc = 0;
David Teigland814d6862006-06-26 00:27:31 -0700541
542 if (!input) {
543 *argvp = NULL;
544 return 0;
545 }
546
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 argv = realloc_argv(&array_size, argv);
548 if (!argv)
549 return -ENOMEM;
550
551 while (1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 /* Skip whitespace */
André Goddard Rosae7d28602009-12-14 18:01:06 -0800553 start = skip_spaces(end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
555 if (!*start)
556 break; /* success, we hit the end */
557
558 /* 'out' is used to remove any back-quotes */
559 end = out = start;
560 while (*end) {
561 /* Everything apart from '\0' can be quoted */
562 if (*end == '\\' && *(end + 1)) {
563 *out++ = *(end + 1);
564 end += 2;
565 continue;
566 }
567
568 if (isspace(*end))
569 break; /* end of token */
570
571 *out++ = *end++;
572 }
573
574 /* have we already filled the array ? */
575 if ((*argc + 1) > array_size) {
576 argv = realloc_argv(&array_size, argv);
577 if (!argv)
578 return -ENOMEM;
579 }
580
581 /* we know this is whitespace */
582 if (*end)
583 end++;
584
585 /* terminate the string and put it in the array */
586 *out = '\0';
587 argv[*argc] = start;
588 (*argc)++;
589 }
590
591 *argvp = argv;
592 return 0;
593}
594
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100595/*
596 * Impose necessary and sufficient conditions on a devices's table such
597 * that any incoming bio which respects its logical_block_size can be
598 * processed successfully. If it falls across the boundary between
599 * two or more targets, the size of each piece it gets split into must
600 * be compatible with the logical_block_size of the target processing it.
601 */
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100602static int validate_hardware_logical_block_alignment(struct dm_table *table,
603 struct queue_limits *limits)
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100604{
605 /*
606 * This function uses arithmetic modulo the logical_block_size
607 * (in units of 512-byte sectors).
608 */
609 unsigned short device_logical_block_size_sects =
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100610 limits->logical_block_size >> SECTOR_SHIFT;
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100611
612 /*
613 * Offset of the start of the next table entry, mod logical_block_size.
614 */
615 unsigned short next_target_start = 0;
616
617 /*
618 * Given an aligned bio that extends beyond the end of a
619 * target, how many sectors must the next target handle?
620 */
621 unsigned short remaining = 0;
622
623 struct dm_target *uninitialized_var(ti);
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100624 struct queue_limits ti_limits;
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100625 unsigned i = 0;
626
627 /*
628 * Check each entry in the table in turn.
629 */
630 while (i < dm_table_get_num_targets(table)) {
631 ti = dm_table_get_target(table, i++);
632
Martin K. Petersenb1bd0552012-01-11 16:27:11 +0100633 blk_set_stacking_limits(&ti_limits);
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100634
635 /* combine all target devices' limits */
636 if (ti->type->iterate_devices)
637 ti->type->iterate_devices(ti, dm_set_device_limits,
638 &ti_limits);
639
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100640 /*
641 * If the remaining sectors fall entirely within this
642 * table entry are they compatible with its logical_block_size?
643 */
644 if (remaining < ti->len &&
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100645 remaining & ((ti_limits.logical_block_size >>
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100646 SECTOR_SHIFT) - 1))
647 break; /* Error */
648
649 next_target_start =
650 (unsigned short) ((next_target_start + ti->len) &
651 (device_logical_block_size_sects - 1));
652 remaining = next_target_start ?
653 device_logical_block_size_sects - next_target_start : 0;
654 }
655
656 if (remaining) {
657 DMWARN("%s: table line %u (start sect %llu len %llu) "
Mike Snitzera963a952009-09-04 20:40:24 +0100658 "not aligned to h/w logical block size %u",
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100659 dm_device_name(table->md), i,
660 (unsigned long long) ti->begin,
661 (unsigned long long) ti->len,
Mike Snitzer754c5fc2009-06-22 10:12:34 +0100662 limits->logical_block_size);
Mike Snitzerbe6d4302009-06-22 10:12:31 +0100663 return -EINVAL;
664 }
665
666 return 0;
667}
668
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669int dm_table_add_target(struct dm_table *t, const char *type,
670 sector_t start, sector_t len, char *params)
671{
672 int r = -EINVAL, argc;
673 char **argv;
674 struct dm_target *tgt;
675
Alasdair G Kergon3791e2f2011-10-31 20:19:00 +0000676 if (t->singleton) {
677 DMERR("%s: target type %s must appear alone in table",
678 dm_device_name(t->md), t->targets->type->name);
679 return -EINVAL;
680 }
681
Mikulas Patocka57a2f232013-11-22 19:51:39 -0500682 BUG_ON(t->num_targets >= t->num_allocated);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
684 tgt = t->targets + t->num_targets;
685 memset(tgt, 0, sizeof(*tgt));
686
687 if (!len) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700688 DMERR("%s: zero-length target", dm_device_name(t->md));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 return -EINVAL;
690 }
691
692 tgt->type = dm_get_target_type(type);
693 if (!tgt->type) {
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700694 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
695 type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 return -EINVAL;
697 }
698
Alasdair G Kergon3791e2f2011-10-31 20:19:00 +0000699 if (dm_target_needs_singleton(tgt->type)) {
700 if (t->num_targets) {
701 DMERR("%s: target type %s must appear alone in table",
702 dm_device_name(t->md), type);
703 return -EINVAL;
704 }
705 t->singleton = 1;
706 }
707
Alasdair G Kergoncc6cbe12011-10-31 20:19:02 +0000708 if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
709 DMERR("%s: target type %s may not be included in read-only tables",
710 dm_device_name(t->md), type);
711 return -EINVAL;
712 }
713
Alasdair G Kergon36a04562011-10-31 20:19:04 +0000714 if (t->immutable_target_type) {
715 if (t->immutable_target_type != tgt->type) {
716 DMERR("%s: immutable target type %s cannot be mixed with other target types",
717 dm_device_name(t->md), t->immutable_target_type->name);
718 return -EINVAL;
719 }
720 } else if (dm_target_is_immutable(tgt->type)) {
721 if (t->num_targets) {
722 DMERR("%s: immutable target type %s cannot be mixed with other target types",
723 dm_device_name(t->md), tgt->type->name);
724 return -EINVAL;
725 }
726 t->immutable_target_type = tgt->type;
727 }
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 tgt->table = t;
730 tgt->begin = start;
731 tgt->len = len;
732 tgt->error = "Unknown error";
733
734 /*
735 * Does this target adjoin the previous one ?
736 */
737 if (!adjoin(t, tgt)) {
738 tgt->error = "Gap in table";
739 r = -EINVAL;
740 goto bad;
741 }
742
743 r = dm_split_args(&argc, &argv, params);
744 if (r) {
745 tgt->error = "couldn't split parameters (insufficient memory)";
746 goto bad;
747 }
748
749 r = tgt->type->ctr(tgt, argc, argv);
750 kfree(argv);
751 if (r)
752 goto bad;
753
754 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
755
Alasdair G Kergon55a62ee2013-03-01 22:45:47 +0000756 if (!tgt->num_discard_bios && tgt->discards_supported)
757 DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
Mike Snitzer936688d2011-08-02 12:32:01 +0100758 dm_device_name(t->md), type);
Mike Snitzer5ae89a82010-08-12 04:14:08 +0100759
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 return 0;
761
762 bad:
Alasdair G Kergon72d94862006-06-26 00:27:35 -0700763 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 dm_put_target_type(tgt->type);
765 return r;
766}
767
Mike Snitzer498f0102011-08-02 12:32:04 +0100768/*
769 * Target argument parsing helpers.
770 */
771static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
772 unsigned *value, char **error, unsigned grouped)
773{
774 const char *arg_str = dm_shift_arg(arg_set);
Mikulas Patocka31998ef2012-03-28 18:41:26 +0100775 char dummy;
Mike Snitzer498f0102011-08-02 12:32:04 +0100776
777 if (!arg_str ||
Mikulas Patocka31998ef2012-03-28 18:41:26 +0100778 (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
Mike Snitzer498f0102011-08-02 12:32:04 +0100779 (*value < arg->min) ||
780 (*value > arg->max) ||
781 (grouped && arg_set->argc < *value)) {
782 *error = arg->error;
783 return -EINVAL;
784 }
785
786 return 0;
787}
788
789int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
790 unsigned *value, char **error)
791{
792 return validate_next_arg(arg, arg_set, value, error, 0);
793}
794EXPORT_SYMBOL(dm_read_arg);
795
796int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set,
797 unsigned *value, char **error)
798{
799 return validate_next_arg(arg, arg_set, value, error, 1);
800}
801EXPORT_SYMBOL(dm_read_arg_group);
802
803const char *dm_shift_arg(struct dm_arg_set *as)
804{
805 char *r;
806
807 if (as->argc) {
808 as->argc--;
809 r = *as->argv;
810 as->argv++;
811 return r;
812 }
813
814 return NULL;
815}
816EXPORT_SYMBOL(dm_shift_arg);
817
818void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
819{
820 BUG_ON(as->argc < num_args);
821 as->argc -= num_args;
822 as->argv += num_args;
823}
824EXPORT_SYMBOL(dm_consume_args);
825
Will Drewry26803b92010-08-12 04:14:03 +0100826static int dm_table_set_type(struct dm_table *t)
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100827{
828 unsigned i;
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400829 unsigned bio_based = 0, request_based = 0, hybrid = 0;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100830 struct dm_target *tgt;
831 struct dm_dev_internal *dd;
832 struct list_head *devices;
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400833 unsigned live_md_type;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100834
835 for (i = 0; i < t->num_targets; i++) {
836 tgt = t->targets + i;
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400837 if (dm_target_hybrid(tgt))
838 hybrid = 1;
839 else if (dm_target_request_based(tgt))
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100840 request_based = 1;
841 else
842 bio_based = 1;
843
844 if (bio_based && request_based) {
845 DMWARN("Inconsistent table: different target types"
846 " can't be mixed up");
847 return -EINVAL;
848 }
849 }
850
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400851 if (hybrid && !bio_based && !request_based) {
852 /*
853 * The targets can work either way.
854 * Determine the type from the live device.
855 * Default to bio-based if device is new.
856 */
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400857 live_md_type = dm_get_md_type(t->md);
Mike Snitzer169e2cc2013-08-22 18:21:38 -0400858 if (live_md_type == DM_TYPE_REQUEST_BASED)
859 request_based = 1;
860 else
861 bio_based = 1;
862 }
863
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100864 if (bio_based) {
865 /* We must use this table as bio-based */
866 t->type = DM_TYPE_BIO_BASED;
867 return 0;
868 }
869
870 BUG_ON(!request_based); /* No targets in this table */
871
872 /* Non-request-stackable devices can't be used for request-based dm */
873 devices = dm_table_get_devices(t);
874 list_for_each_entry(dd, devices, list) {
Benjamin Marzinski86f11522014-08-13 13:53:43 -0500875 if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev->bdev))) {
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100876 DMWARN("table load rejected: including"
877 " non-request-stackable devices");
878 return -EINVAL;
879 }
880 }
881
882 /*
883 * Request-based dm supports only tables that have a single target now.
884 * To support multiple targets, request splitting support is needed,
885 * and that needs lots of changes in the block-layer.
886 * (e.g. request completion process for partial completion.)
887 */
888 if (t->num_targets > 1) {
889 DMWARN("Request-based dm doesn't support multiple targets yet");
890 return -EINVAL;
891 }
892
893 t->type = DM_TYPE_REQUEST_BASED;
894
895 return 0;
896}
897
898unsigned dm_table_get_type(struct dm_table *t)
899{
900 return t->type;
901}
902
Alasdair G Kergon36a04562011-10-31 20:19:04 +0000903struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
904{
905 return t->immutable_target_type;
906}
907
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100908bool dm_table_request_based(struct dm_table *t)
909{
910 return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED;
911}
912
Mikulas Patocka473c36d2014-02-13 13:43:32 -0500913static int dm_table_alloc_md_mempools(struct dm_table *t)
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100914{
915 unsigned type = dm_table_get_type(t);
Mikulas Patockac0820cf2012-12-21 20:23:38 +0000916 unsigned per_bio_data_size = 0;
917 struct dm_target *tgt;
918 unsigned i;
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100919
920 if (unlikely(type == DM_TYPE_NONE)) {
921 DMWARN("no table type is set, can't allocate mempools");
922 return -EINVAL;
923 }
924
Mikulas Patockac0820cf2012-12-21 20:23:38 +0000925 if (type == DM_TYPE_BIO_BASED)
926 for (i = 0; i < t->num_targets; i++) {
927 tgt = t->targets + i;
928 per_bio_data_size = max(per_bio_data_size, tgt->per_bio_data_size);
929 }
930
931 t->mempools = dm_alloc_md_mempools(type, t->integrity_supported, per_bio_data_size);
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +0100932 if (!t->mempools)
933 return -ENOMEM;
934
935 return 0;
936}
937
938void dm_table_free_md_mempools(struct dm_table *t)
939{
940 dm_free_md_mempools(t->mempools);
941 t->mempools = NULL;
942}
943
944struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
945{
946 return t->mempools;
947}
948
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949static int setup_indexes(struct dm_table *t)
950{
951 int i;
952 unsigned int total = 0;
953 sector_t *indexes;
954
955 /* allocate the space for *all* the indexes */
956 for (i = t->depth - 2; i >= 0; i--) {
957 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
958 total += t->counts[i];
959 }
960
961 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
962 if (!indexes)
963 return -ENOMEM;
964
965 /* set up internal nodes, bottom-up */
Jun'ichi Nomura82d601d2008-02-08 02:10:04 +0000966 for (i = t->depth - 2; i >= 0; i--) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 t->index[i] = indexes;
968 indexes += (KEYS_PER_NODE * t->counts[i]);
969 setup_btree_index(i, t);
970 }
971
972 return 0;
973}
974
975/*
976 * Builds the btree to index the map.
977 */
Will Drewry26803b92010-08-12 04:14:03 +0100978static int dm_table_build_index(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979{
980 int r = 0;
981 unsigned int leaf_nodes;
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 /* how many indexes will the btree have ? */
984 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
985 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
986
987 /* leaf layer has already been set up */
988 t->counts[t->depth - 1] = leaf_nodes;
989 t->index[t->depth - 1] = t->highs;
990
991 if (t->depth >= 2)
992 r = setup_indexes(t);
993
994 return r;
995}
996
Will Drewry26803b92010-08-12 04:14:03 +0100997/*
Mike Snitzera63a5cf2011-04-01 21:02:31 +0200998 * Get a disk whose integrity profile reflects the table's profile.
999 * If %match_all is true, all devices' profiles must match.
1000 * If %match_all is false, all devices must at least have an
1001 * allocated integrity profile; but uninitialized is ok.
1002 * Returns NULL if integrity support was inconsistent or unavailable.
1003 */
1004static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t,
1005 bool match_all)
1006{
1007 struct list_head *devices = dm_table_get_devices(t);
1008 struct dm_dev_internal *dd = NULL;
1009 struct gendisk *prev_disk = NULL, *template_disk = NULL;
1010
1011 list_for_each_entry(dd, devices, list) {
Benjamin Marzinski86f11522014-08-13 13:53:43 -05001012 template_disk = dd->dm_dev->bdev->bd_disk;
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001013 if (!blk_get_integrity(template_disk))
1014 goto no_integrity;
1015 if (!match_all && !blk_integrity_is_initialized(template_disk))
1016 continue; /* skip uninitialized profiles */
1017 else if (prev_disk &&
1018 blk_integrity_compare(prev_disk, template_disk) < 0)
1019 goto no_integrity;
1020 prev_disk = template_disk;
1021 }
1022
1023 return template_disk;
1024
1025no_integrity:
1026 if (prev_disk)
1027 DMWARN("%s: integrity not set: %s and %s profile mismatch",
1028 dm_device_name(t->md),
1029 prev_disk->disk_name,
1030 template_disk->disk_name);
1031 return NULL;
1032}
1033
1034/*
Will Drewry26803b92010-08-12 04:14:03 +01001035 * Register the mapped device for blk_integrity support if
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001036 * the underlying devices have an integrity profile. But all devices
1037 * may not have matching profiles (checking all devices isn't reliable
1038 * during table load because this table may use other DM device(s) which
1039 * must be resumed before they will have an initialized integity profile).
1040 * Stacked DM devices force a 2 stage integrity profile validation:
1041 * 1 - during load, validate all initialized integrity profiles match
1042 * 2 - during resume, validate all integrity profiles match
Will Drewry26803b92010-08-12 04:14:03 +01001043 */
1044static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md)
1045{
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001046 struct gendisk *template_disk = NULL;
Will Drewry26803b92010-08-12 04:14:03 +01001047
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001048 template_disk = dm_table_get_integrity_disk(t, false);
1049 if (!template_disk)
1050 return 0;
Will Drewry26803b92010-08-12 04:14:03 +01001051
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001052 if (!blk_integrity_is_initialized(dm_disk(md))) {
1053 t->integrity_supported = 1;
1054 return blk_integrity_register(dm_disk(md), NULL);
1055 }
1056
1057 /*
1058 * If DM device already has an initalized integrity
1059 * profile the new profile should not conflict.
1060 */
1061 if (blk_integrity_is_initialized(template_disk) &&
1062 blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1063 DMWARN("%s: conflict with existing integrity profile: "
1064 "%s profile mismatch",
1065 dm_device_name(t->md),
1066 template_disk->disk_name);
1067 return 1;
1068 }
1069
1070 /* Preserve existing initialized integrity profile */
1071 t->integrity_supported = 1;
Will Drewry26803b92010-08-12 04:14:03 +01001072 return 0;
1073}
1074
1075/*
1076 * Prepares the table for use by building the indices,
1077 * setting the type, and allocating mempools.
1078 */
1079int dm_table_complete(struct dm_table *t)
1080{
1081 int r;
1082
1083 r = dm_table_set_type(t);
1084 if (r) {
1085 DMERR("unable to set table type");
1086 return r;
1087 }
1088
1089 r = dm_table_build_index(t);
1090 if (r) {
1091 DMERR("unable to build btrees");
1092 return r;
1093 }
1094
1095 r = dm_table_prealloc_integrity(t, t->md);
1096 if (r) {
1097 DMERR("could not register integrity profile.");
1098 return r;
1099 }
1100
1101 r = dm_table_alloc_md_mempools(t);
1102 if (r)
1103 DMERR("unable to allocate mempools");
1104
1105 return r;
1106}
1107
Arjan van de Ven48c9c272006-03-27 01:18:20 -08001108static DEFINE_MUTEX(_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109void dm_table_event_callback(struct dm_table *t,
1110 void (*fn)(void *), void *context)
1111{
Arjan van de Ven48c9c272006-03-27 01:18:20 -08001112 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 t->event_fn = fn;
1114 t->event_context = context;
Arjan van de Ven48c9c272006-03-27 01:18:20 -08001115 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116}
1117
1118void dm_table_event(struct dm_table *t)
1119{
1120 /*
1121 * You can no longer call dm_table_event() from interrupt
1122 * context, use a bottom half instead.
1123 */
1124 BUG_ON(in_interrupt());
1125
Arjan van de Ven48c9c272006-03-27 01:18:20 -08001126 mutex_lock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 if (t->event_fn)
1128 t->event_fn(t->event_context);
Arjan van de Ven48c9c272006-03-27 01:18:20 -08001129 mutex_unlock(&_event_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130}
Mike Snitzer08649012011-08-02 12:32:04 +01001131EXPORT_SYMBOL(dm_table_event);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
1133sector_t dm_table_get_size(struct dm_table *t)
1134{
1135 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1136}
Mike Snitzer08649012011-08-02 12:32:04 +01001137EXPORT_SYMBOL(dm_table_get_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
1139struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1140{
Milan Broz14353532006-06-26 00:27:27 -07001141 if (index >= t->num_targets)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 return NULL;
1143
1144 return t->targets + index;
1145}
1146
1147/*
1148 * Search the btree for the correct target.
Jun'ichi Nomura512875b2007-12-13 14:15:25 +00001149 *
1150 * Caller should check returned pointer with dm_target_is_valid()
1151 * to trap I/O beyond end of device.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 */
1153struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1154{
1155 unsigned int l, n = 0, k = 0;
1156 sector_t *node;
1157
1158 for (l = 0; l < t->depth; l++) {
1159 n = get_child(n, k);
1160 node = get_node(t, l, n);
1161
1162 for (k = 0; k < KEYS_PER_NODE; k++)
1163 if (node[k] >= sector)
1164 break;
1165 }
1166
1167 return &t->targets[(KEYS_PER_NODE * n) + k];
1168}
1169
Mike Snitzer3ae70652012-09-26 23:45:45 +01001170static int count_device(struct dm_target *ti, struct dm_dev *dev,
1171 sector_t start, sector_t len, void *data)
1172{
1173 unsigned *num_devices = data;
1174
1175 (*num_devices)++;
1176
1177 return 0;
1178}
1179
1180/*
1181 * Check whether a table has no data devices attached using each
1182 * target's iterate_devices method.
1183 * Returns false if the result is unknown because a target doesn't
1184 * support iterate_devices.
1185 */
1186bool dm_table_has_no_data_devices(struct dm_table *table)
1187{
1188 struct dm_target *uninitialized_var(ti);
1189 unsigned i = 0, num_devices = 0;
1190
1191 while (i < dm_table_get_num_targets(table)) {
1192 ti = dm_table_get_target(table, i++);
1193
1194 if (!ti->type->iterate_devices)
1195 return false;
1196
1197 ti->type->iterate_devices(ti, count_device, &num_devices);
1198 if (num_devices)
1199 return false;
1200 }
1201
1202 return true;
1203}
1204
Martin K. Petersen9c470082009-04-09 00:27:12 +01001205/*
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001206 * Establish the new table's queue_limits and validate them.
1207 */
1208int dm_calculate_queue_limits(struct dm_table *table,
1209 struct queue_limits *limits)
1210{
1211 struct dm_target *uninitialized_var(ti);
1212 struct queue_limits ti_limits;
1213 unsigned i = 0;
1214
Martin K. Petersenb1bd0552012-01-11 16:27:11 +01001215 blk_set_stacking_limits(limits);
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001216
1217 while (i < dm_table_get_num_targets(table)) {
Martin K. Petersenb1bd0552012-01-11 16:27:11 +01001218 blk_set_stacking_limits(&ti_limits);
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001219
1220 ti = dm_table_get_target(table, i++);
1221
1222 if (!ti->type->iterate_devices)
1223 goto combine_limits;
1224
1225 /*
1226 * Combine queue limits of all the devices this target uses.
1227 */
1228 ti->type->iterate_devices(ti, dm_set_device_limits,
1229 &ti_limits);
1230
Mike Snitzer40bea432009-09-04 20:40:25 +01001231 /* Set I/O hints portion of queue limits */
1232 if (ti->type->io_hints)
1233 ti->type->io_hints(ti, &ti_limits);
1234
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001235 /*
1236 * Check each device area is consistent with the target's
1237 * overall queue limits.
1238 */
Mikulas Patockaf6a1ed12009-09-04 20:40:22 +01001239 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1240 &ti_limits))
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001241 return -EINVAL;
1242
1243combine_limits:
1244 /*
1245 * Merge this target's queue limits into the overall limits
1246 * for the table.
1247 */
1248 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
Martin K. Petersenb27d7f12010-01-11 03:21:50 -05001249 DMWARN("%s: adding target device "
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001250 "(start sect %llu len %llu) "
Martin K. Petersenb27d7f12010-01-11 03:21:50 -05001251 "caused an alignment inconsistency",
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001252 dm_device_name(table->md),
1253 (unsigned long long) ti->begin,
1254 (unsigned long long) ti->len);
1255 }
1256
1257 return validate_hardware_logical_block_alignment(table, limits);
1258}
1259
1260/*
Martin K. Petersen9c470082009-04-09 00:27:12 +01001261 * Set the integrity profile for this device if all devices used have
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001262 * matching profiles. We're quite deep in the resume path but still
1263 * don't know if all devices (particularly DM devices this device
1264 * may be stacked on) have matching profiles. Even if the profiles
1265 * don't match we have no way to fail (to resume) at this point.
Martin K. Petersen9c470082009-04-09 00:27:12 +01001266 */
1267static void dm_table_set_integrity(struct dm_table *t)
1268{
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001269 struct gendisk *template_disk = NULL;
Martin K. Petersen9c470082009-04-09 00:27:12 +01001270
1271 if (!blk_get_integrity(dm_disk(t->md)))
1272 return;
1273
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001274 template_disk = dm_table_get_integrity_disk(t, true);
Mike Snitzer876fbba2011-09-25 23:26:17 +01001275 if (template_disk)
1276 blk_integrity_register(dm_disk(t->md),
1277 blk_get_integrity(template_disk));
1278 else if (blk_integrity_is_initialized(dm_disk(t->md)))
Mike Snitzera63a5cf2011-04-01 21:02:31 +02001279 DMWARN("%s: device no longer has a valid integrity profile",
1280 dm_device_name(t->md));
Mike Snitzer876fbba2011-09-25 23:26:17 +01001281 else
1282 DMWARN("%s: unable to establish an integrity profile",
1283 dm_device_name(t->md));
Martin K. Petersen9c470082009-04-09 00:27:12 +01001284}
1285
Mike Snitzered8b7522011-08-02 12:32:08 +01001286static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1287 sector_t start, sector_t len, void *data)
1288{
1289 unsigned flush = (*(unsigned *)data);
1290 struct request_queue *q = bdev_get_queue(dev->bdev);
1291
1292 return q && (q->flush_flags & flush);
1293}
1294
1295static bool dm_table_supports_flush(struct dm_table *t, unsigned flush)
1296{
1297 struct dm_target *ti;
1298 unsigned i = 0;
1299
1300 /*
1301 * Require at least one underlying device to support flushes.
1302 * t->devices includes internal dm devices such as mirror logs
1303 * so we need to use iterate_devices here, which targets
1304 * supporting flushes must provide.
1305 */
1306 while (i < dm_table_get_num_targets(t)) {
1307 ti = dm_table_get_target(t, i++);
1308
Alasdair G Kergon55a62ee2013-03-01 22:45:47 +00001309 if (!ti->num_flush_bios)
Mike Snitzered8b7522011-08-02 12:32:08 +01001310 continue;
1311
Joe Thornber0e9c24e2012-07-27 15:08:07 +01001312 if (ti->flush_supported)
1313 return 1;
1314
Mike Snitzered8b7522011-08-02 12:32:08 +01001315 if (ti->type->iterate_devices &&
1316 ti->type->iterate_devices(ti, device_flush_capable, &flush))
1317 return 1;
1318 }
1319
1320 return 0;
1321}
1322
Milan Broz983c7db2011-09-25 23:26:21 +01001323static bool dm_table_discard_zeroes_data(struct dm_table *t)
1324{
1325 struct dm_target *ti;
1326 unsigned i = 0;
1327
1328 /* Ensure that all targets supports discard_zeroes_data. */
1329 while (i < dm_table_get_num_targets(t)) {
1330 ti = dm_table_get_target(t, i++);
1331
1332 if (ti->discard_zeroes_data_unsupported)
1333 return 0;
1334 }
1335
1336 return 1;
1337}
1338
Mandeep Singh Baines4693c9662011-10-31 20:18:50 +00001339static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
1340 sector_t start, sector_t len, void *data)
1341{
1342 struct request_queue *q = bdev_get_queue(dev->bdev);
1343
1344 return q && blk_queue_nonrot(q);
1345}
1346
Milan Brozc3c4555e2012-09-26 23:45:43 +01001347static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1348 sector_t start, sector_t len, void *data)
1349{
1350 struct request_queue *q = bdev_get_queue(dev->bdev);
1351
1352 return q && !blk_queue_add_random(q);
1353}
1354
Jeff Moyer200612e2014-08-08 11:03:41 -04001355static int queue_supports_sg_merge(struct dm_target *ti, struct dm_dev *dev,
1356 sector_t start, sector_t len, void *data)
1357{
1358 struct request_queue *q = bdev_get_queue(dev->bdev);
1359
1360 return q && !test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags);
1361}
1362
Milan Brozc3c4555e2012-09-26 23:45:43 +01001363static bool dm_table_all_devices_attribute(struct dm_table *t,
1364 iterate_devices_callout_fn func)
Mandeep Singh Baines4693c9662011-10-31 20:18:50 +00001365{
1366 struct dm_target *ti;
1367 unsigned i = 0;
1368
Mandeep Singh Baines4693c9662011-10-31 20:18:50 +00001369 while (i < dm_table_get_num_targets(t)) {
1370 ti = dm_table_get_target(t, i++);
1371
1372 if (!ti->type->iterate_devices ||
Milan Brozc3c4555e2012-09-26 23:45:43 +01001373 !ti->type->iterate_devices(ti, func, NULL))
Mandeep Singh Baines4693c9662011-10-31 20:18:50 +00001374 return 0;
1375 }
1376
1377 return 1;
1378}
1379
Mike Snitzerd54eaa52012-12-21 20:23:36 +00001380static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1381 sector_t start, sector_t len, void *data)
1382{
1383 struct request_queue *q = bdev_get_queue(dev->bdev);
1384
1385 return q && !q->limits.max_write_same_sectors;
1386}
1387
1388static bool dm_table_supports_write_same(struct dm_table *t)
1389{
1390 struct dm_target *ti;
1391 unsigned i = 0;
1392
1393 while (i < dm_table_get_num_targets(t)) {
1394 ti = dm_table_get_target(t, i++);
1395
Alasdair G Kergon55a62ee2013-03-01 22:45:47 +00001396 if (!ti->num_write_same_bios)
Mike Snitzerd54eaa52012-12-21 20:23:36 +00001397 return false;
1398
1399 if (!ti->type->iterate_devices ||
Mike Snitzerdc019b22013-05-10 14:37:16 +01001400 ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
Mike Snitzerd54eaa52012-12-21 20:23:36 +00001401 return false;
1402 }
1403
1404 return true;
1405}
1406
Mikulas Patockaa7ffb6a2014-07-10 12:23:07 -04001407static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1408 sector_t start, sector_t len, void *data)
1409{
1410 struct request_queue *q = bdev_get_queue(dev->bdev);
1411
1412 return q && blk_queue_discard(q);
1413}
1414
1415static bool dm_table_supports_discards(struct dm_table *t)
1416{
1417 struct dm_target *ti;
1418 unsigned i = 0;
1419
1420 /*
1421 * Unless any target used by the table set discards_supported,
1422 * require at least one underlying device to support discards.
1423 * t->devices includes internal dm devices such as mirror logs
1424 * so we need to use iterate_devices here, which targets
1425 * supporting discard selectively must provide.
1426 */
1427 while (i < dm_table_get_num_targets(t)) {
1428 ti = dm_table_get_target(t, i++);
1429
1430 if (!ti->num_discard_bios)
1431 continue;
1432
1433 if (ti->discards_supported)
1434 return 1;
1435
1436 if (ti->type->iterate_devices &&
1437 ti->type->iterate_devices(ti, device_discard_capable, NULL))
1438 return 1;
1439 }
1440
1441 return 0;
1442}
1443
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001444void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1445 struct queue_limits *limits)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446{
Mike Snitzered8b7522011-08-02 12:32:08 +01001447 unsigned flush = 0;
1448
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 /*
Mike Snitzer11977642009-06-22 10:12:32 +01001450 * Copy table's limits to the DM device's request_queue
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 */
Mike Snitzer754c5fc2009-06-22 10:12:34 +01001452 q->limits = *limits;
Jens Axboec9a3f6d2008-04-29 19:12:35 +02001453
Mike Snitzer5ae89a82010-08-12 04:14:08 +01001454 if (!dm_table_supports_discards(t))
1455 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q);
1456 else
1457 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);
1458
Mike Snitzered8b7522011-08-02 12:32:08 +01001459 if (dm_table_supports_flush(t, REQ_FLUSH)) {
1460 flush |= REQ_FLUSH;
1461 if (dm_table_supports_flush(t, REQ_FUA))
1462 flush |= REQ_FUA;
1463 }
1464 blk_queue_flush(q, flush);
1465
Milan Broz983c7db2011-09-25 23:26:21 +01001466 if (!dm_table_discard_zeroes_data(t))
1467 q->limits.discard_zeroes_data = 0;
1468
Milan Brozc3c4555e2012-09-26 23:45:43 +01001469 /* Ensure that all underlying devices are non-rotational. */
1470 if (dm_table_all_devices_attribute(t, device_is_nonrot))
Mandeep Singh Baines4693c9662011-10-31 20:18:50 +00001471 queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
1472 else
1473 queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q);
1474
Mike Snitzerd54eaa52012-12-21 20:23:36 +00001475 if (!dm_table_supports_write_same(t))
1476 q->limits.max_write_same_sectors = 0;
Mike Snitzerc1a94672012-12-21 20:23:30 +00001477
Jeff Moyer200612e2014-08-08 11:03:41 -04001478 if (dm_table_all_devices_attribute(t, queue_supports_sg_merge))
1479 queue_flag_clear_unlocked(QUEUE_FLAG_NO_SG_MERGE, q);
1480 else
1481 queue_flag_set_unlocked(QUEUE_FLAG_NO_SG_MERGE, q);
1482
Martin K. Petersen9c470082009-04-09 00:27:12 +01001483 dm_table_set_integrity(t);
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +01001484
1485 /*
Milan Brozc3c4555e2012-09-26 23:45:43 +01001486 * Determine whether or not this queue's I/O timings contribute
1487 * to the entropy pool, Only request-based targets use this.
1488 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1489 * have it set.
1490 */
1491 if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
1492 queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, q);
1493
1494 /*
Kiyoshi Uedae6ee8c02009-06-22 10:12:36 +01001495 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1496 * visible to other CPUs because, once the flag is set, incoming bios
1497 * are processed by request-based dm, which refers to the queue
1498 * settings.
1499 * Until the flag set, bios are passed to bio-based dm and queued to
1500 * md->deferred where queue settings are not needed yet.
1501 * Those bios are passed to request-based dm at the resume time.
1502 */
1503 smp_mb();
1504 if (dm_table_request_based(t))
1505 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506}
1507
1508unsigned int dm_table_get_num_targets(struct dm_table *t)
1509{
1510 return t->num_targets;
1511}
1512
1513struct list_head *dm_table_get_devices(struct dm_table *t)
1514{
1515 return &t->devices;
1516}
1517
Al Viroaeb5d722008-09-02 15:28:45 -04001518fmode_t dm_table_get_mode(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519{
1520 return t->mode;
1521}
Mike Snitzer08649012011-08-02 12:32:04 +01001522EXPORT_SYMBOL(dm_table_get_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
1524static void suspend_targets(struct dm_table *t, unsigned postsuspend)
1525{
1526 int i = t->num_targets;
1527 struct dm_target *ti = t->targets;
1528
1529 while (i--) {
1530 if (postsuspend) {
1531 if (ti->type->postsuspend)
1532 ti->type->postsuspend(ti);
1533 } else if (ti->type->presuspend)
1534 ti->type->presuspend(ti);
1535
1536 ti++;
1537 }
1538}
1539
1540void dm_table_presuspend_targets(struct dm_table *t)
1541{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -07001542 if (!t)
1543 return;
1544
Adrian Bunke8488d02008-04-24 22:10:51 +01001545 suspend_targets(t, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546}
1547
1548void dm_table_postsuspend_targets(struct dm_table *t)
1549{
Alasdair G Kergoncf222b32005-07-28 21:15:57 -07001550 if (!t)
1551 return;
1552
Adrian Bunke8488d02008-04-24 22:10:51 +01001553 suspend_targets(t, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554}
1555
Milan Broz8757b772006-10-03 01:15:36 -07001556int dm_table_resume_targets(struct dm_table *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557{
Milan Broz8757b772006-10-03 01:15:36 -07001558 int i, r = 0;
1559
1560 for (i = 0; i < t->num_targets; i++) {
1561 struct dm_target *ti = t->targets + i;
1562
1563 if (!ti->type->preresume)
1564 continue;
1565
1566 r = ti->type->preresume(ti);
Mike Snitzer7833b082013-10-24 14:10:29 -04001567 if (r) {
1568 DMERR("%s: %s: preresume failed, error = %d",
1569 dm_device_name(t->md), ti->type->name, r);
Milan Broz8757b772006-10-03 01:15:36 -07001570 return r;
Mike Snitzer7833b082013-10-24 14:10:29 -04001571 }
Milan Broz8757b772006-10-03 01:15:36 -07001572 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
1574 for (i = 0; i < t->num_targets; i++) {
1575 struct dm_target *ti = t->targets + i;
1576
1577 if (ti->type->resume)
1578 ti->type->resume(ti);
1579 }
Milan Broz8757b772006-10-03 01:15:36 -07001580
1581 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582}
1583
NeilBrown9d357b02011-01-13 20:00:01 +00001584void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
1585{
1586 list_add(&cb->list, &t->target_callbacks);
1587}
1588EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
1589
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590int dm_table_any_congested(struct dm_table *t, int bdi_bits)
1591{
Mikulas Patocka82b15192008-10-10 13:37:09 +01001592 struct dm_dev_internal *dd;
Paul Jimenezafb24522008-02-08 02:09:59 +00001593 struct list_head *devices = dm_table_get_devices(t);
NeilBrown9d357b02011-01-13 20:00:01 +00001594 struct dm_target_callbacks *cb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 int r = 0;
1596
Paul Jimenezafb24522008-02-08 02:09:59 +00001597 list_for_each_entry(dd, devices, list) {
Benjamin Marzinski86f11522014-08-13 13:53:43 -05001598 struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev);
Alasdair G Kergon0c2322e2008-10-10 13:37:13 +01001599 char b[BDEVNAME_SIZE];
1600
1601 if (likely(q))
1602 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
1603 else
1604 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1605 dm_device_name(t->md),
Benjamin Marzinski86f11522014-08-13 13:53:43 -05001606 bdevname(dd->dm_dev->bdev, b));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 }
1608
NeilBrown9d357b02011-01-13 20:00:01 +00001609 list_for_each_entry(cb, &t->target_callbacks, list)
1610 if (cb->congested_fn)
1611 r |= cb->congested_fn(cb, bdi_bits);
1612
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 return r;
1614}
1615
Kiyoshi Uedacec47e32009-06-22 10:12:35 +01001616int dm_table_any_busy_target(struct dm_table *t)
1617{
1618 unsigned i;
1619 struct dm_target *ti;
1620
1621 for (i = 0; i < t->num_targets; i++) {
1622 ti = t->targets + i;
1623 if (ti->type->busy && ti->type->busy(ti))
1624 return 1;
1625 }
1626
1627 return 0;
1628}
1629
Mike Anderson1134e5a2006-03-27 01:17:54 -08001630struct mapped_device *dm_table_get_md(struct dm_table *t)
1631{
Mike Anderson1134e5a2006-03-27 01:17:54 -08001632 return t->md;
1633}
Mike Snitzer08649012011-08-02 12:32:04 +01001634EXPORT_SYMBOL(dm_table_get_md);
Mike Anderson1134e5a2006-03-27 01:17:54 -08001635
Mike Snitzer9974fa22014-02-28 15:33:43 +01001636void dm_table_run_md_queue_async(struct dm_table *t)
1637{
1638 struct mapped_device *md;
1639 struct request_queue *queue;
1640 unsigned long flags;
1641
1642 if (!dm_table_request_based(t))
1643 return;
1644
1645 md = dm_table_get_md(t);
1646 queue = dm_get_md_queue(md);
1647 if (queue) {
1648 spin_lock_irqsave(queue->queue_lock, flags);
1649 blk_run_queue_async(queue);
1650 spin_unlock_irqrestore(queue->queue_lock, flags);
1651 }
1652}
1653EXPORT_SYMBOL(dm_table_run_md_queue_async);
1654