]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/md/md-linear.c
Merge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[mirror_ubuntu-bionic-kernel.git] / drivers / md / md-linear.c
CommitLineData
1da177e4
LT
1/*
2 linear.c : Multiple Devices driver for Linux
3 Copyright (C) 1994-96 Marc ZYNGIER
4 <zyngier@ufr-info-p7.ibp.fr> or
5 <maz@gloups.fdn.fr>
6
7 Linear mode management functions.
8
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 2, or (at your option)
12 any later version.
f72ffdd6 13
1da177e4
LT
14 You should have received a copy of the GNU General Public License
15 (for example /usr/src/linux/COPYING); if not, write to the Free
f72ffdd6 16 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
1da177e4
LT
17*/
18
bff61975
N
19#include <linux/blkdev.h>
20#include <linux/raid/md_u.h>
bff61975 21#include <linux/seq_file.h>
056075c7 22#include <linux/module.h>
5a0e3ad6 23#include <linux/slab.h>
109e3765 24#include <trace/events/block.h>
43b2e5d8 25#include "md.h"
935fe098 26#include "md-linear.h"
1da177e4 27
1da177e4 28/*
f72ffdd6 29 * find which device holds a particular offset
1da177e4 30 */
a7120771 31static inline struct dev_info *which_dev(struct mddev *mddev, sector_t sector)
1da177e4 32{
aece3d1f 33 int lo, mid, hi;
e849b938 34 struct linear_conf *conf;
1da177e4 35
aece3d1f
SS
36 lo = 0;
37 hi = mddev->raid_disks - 1;
3be260cc 38 conf = mddev->private;
1da177e4 39
aece3d1f
SS
40 /*
41 * Binary Search
42 */
43
44 while (hi > lo) {
45
46 mid = (hi + lo) / 2;
47 if (sector < conf->disks[mid].end_sector)
48 hi = mid;
49 else
50 lo = mid + 1;
51 }
52
53 return conf->disks + lo;
1da177e4
LT
54}
55
03a9e24e 56/*
57 * In linear_congested() conf->raid_disks is used as a copy of
58 * mddev->raid_disks to iterate conf->disks[], because conf->raid_disks
59 * and conf->disks[] are created in linear_conf(), they are always
60 * consitent with each other, but mddev->raid_disks does not.
61 */
5c675f83 62static int linear_congested(struct mddev *mddev, int bits)
26be34dc 63{
e849b938 64 struct linear_conf *conf;
26be34dc
N
65 int i, ret = 0;
66
03a9e24e 67 rcu_read_lock();
68 conf = rcu_dereference(mddev->private);
af11c397 69
03a9e24e 70 for (i = 0; i < conf->raid_disks && !ret ; i++) {
165125e1 71 struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev);
dc3b17cc 72 ret |= bdi_congested(q->backing_dev_info, bits);
26be34dc 73 }
af11c397 74
03a9e24e 75 rcu_read_unlock();
26be34dc
N
76 return ret;
77}
78
fd01b88c 79static sector_t linear_size(struct mddev *mddev, sector_t sectors, int raid_disks)
80c3a6ce 80{
e849b938 81 struct linear_conf *conf;
af11c397 82 sector_t array_sectors;
80c3a6ce 83
3be260cc 84 conf = mddev->private;
80c3a6ce
DW
85 WARN_ONCE(sectors || raid_disks,
86 "%s does not support generic reshape\n", __func__);
af11c397 87 array_sectors = conf->array_sectors;
80c3a6ce 88
af11c397 89 return array_sectors;
80c3a6ce
DW
90}
91
e849b938 92static struct linear_conf *linear_conf(struct mddev *mddev, int raid_disks)
1da177e4 93{
e849b938 94 struct linear_conf *conf;
3cb03002 95 struct md_rdev *rdev;
45d4582f 96 int i, cnt;
f1cad2b6 97 bool discard_supported = false;
1da177e4 98
a7120771 99 conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(struct dev_info),
1da177e4
LT
100 GFP_KERNEL);
101 if (!conf)
7c7546cc
N
102 return NULL;
103
1da177e4 104 cnt = 0;
d6e22150 105 conf->array_sectors = 0;
1da177e4 106
dafb20fa 107 rdev_for_each(rdev, mddev) {
1da177e4 108 int j = rdev->raid_disk;
a7120771 109 struct dev_info *disk = conf->disks + j;
13f2682b 110 sector_t sectors;
1da177e4 111
13864515 112 if (j < 0 || j >= raid_disks || disk->rdev) {
a2e202af
N
113 pr_warn("md/linear:%s: disk numbering problem. Aborting!\n",
114 mdname(mddev));
1da177e4
LT
115 goto out;
116 }
117
118 disk->rdev = rdev;
13f2682b
N
119 if (mddev->chunk_sectors) {
120 sectors = rdev->sectors;
121 sector_div(sectors, mddev->chunk_sectors);
122 rdev->sectors = sectors * mddev->chunk_sectors;
123 }
1da177e4 124
8f6c2e4b
MP
125 disk_stack_limits(mddev->gendisk, rdev->bdev,
126 rdev->data_offset << 9);
1da177e4 127
dd8ac336 128 conf->array_sectors += rdev->sectors;
1da177e4 129 cnt++;
4db7cdc8 130
f1cad2b6
SL
131 if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
132 discard_supported = true;
1da177e4 133 }
7c7546cc 134 if (cnt != raid_disks) {
a2e202af
N
135 pr_warn("md/linear:%s: not enough drives present. Aborting!\n",
136 mdname(mddev));
1da177e4
LT
137 goto out;
138 }
139
f1cad2b6
SL
140 if (!discard_supported)
141 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
142 else
143 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
144
1da177e4 145 /*
45d4582f 146 * Here we calculate the device offsets.
1da177e4 147 */
4db7cdc8
SS
148 conf->disks[0].end_sector = conf->disks[0].rdev->sectors;
149
a778b73f 150 for (i = 1; i < raid_disks; i++)
4db7cdc8
SS
151 conf->disks[i].end_sector =
152 conf->disks[i-1].end_sector +
153 conf->disks[i].rdev->sectors;
15945fee 154
03a9e24e 155 /*
156 * conf->raid_disks is copy of mddev->raid_disks. The reason to
157 * keep a copy of mddev->raid_disks in struct linear_conf is,
158 * mddev->raid_disks may not be consistent with pointers number of
159 * conf->disks[] when it is updated in linear_add() and used to
160 * iterate old conf->disks[] earray in linear_congested().
161 * Here conf->raid_disks is always consitent with number of
162 * pointers in conf->disks[] array, and mddev->private is updated
163 * with rcu_assign_pointer() in linear_addr(), such race can be
164 * avoided.
165 */
166 conf->raid_disks = raid_disks;
167
7c7546cc
N
168 return conf;
169
170out:
171 kfree(conf);
172 return NULL;
173}
174
fd01b88c 175static int linear_run (struct mddev *mddev)
7c7546cc 176{
e849b938 177 struct linear_conf *conf;
98d5561b 178 int ret;
7c7546cc 179
0894cc30
AN
180 if (md_check_no_bitmap(mddev))
181 return -EINVAL;
7c7546cc
N
182 conf = linear_conf(mddev, mddev->raid_disks);
183
184 if (!conf)
185 return 1;
186 mddev->private = conf;
1f403624 187 md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
7c7546cc 188
98d5561b 189 ret = md_integrity_register(mddev);
190 if (ret) {
191 kfree(conf);
192 mddev->private = NULL;
193 }
194 return ret;
7c7546cc 195}
1da177e4 196
fd01b88c 197static int linear_add(struct mddev *mddev, struct md_rdev *rdev)
7c7546cc
N
198{
199 /* Adding a drive to a linear array allows the array to grow.
200 * It is permitted if the new drive has a matching superblock
201 * already on it, with raid_disk equal to raid_disks.
202 * It is achieved by creating a new linear_private_data structure
203 * and swapping it in in-place of the current one.
204 * The current one is never freed until the array is stopped.
205 * This avoids races.
206 */
e849b938 207 struct linear_conf *newconf, *oldconf;
7c7546cc 208
a778b73f 209 if (rdev->saved_raid_disk != mddev->raid_disks)
7c7546cc
N
210 return -EINVAL;
211
a778b73f 212 rdev->raid_disk = rdev->saved_raid_disk;
09cd9270 213 rdev->saved_raid_disk = -1;
a778b73f 214
7c7546cc
N
215 newconf = linear_conf(mddev,mddev->raid_disks+1);
216
217 if (!newconf)
218 return -ENOMEM;
219
03a9e24e 220 /* newconf->raid_disks already keeps a copy of * the increased
221 * value of mddev->raid_disks, WARN_ONCE() is just used to make
222 * sure of this. It is possible that oldconf is still referenced
223 * in linear_congested(), therefore kfree_rcu() is used to free
224 * oldconf until no one uses it anymore.
225 */
3be260cc 226 mddev_suspend(mddev);
d939cdfd
SL
227 oldconf = rcu_dereference_protected(mddev->private,
228 lockdep_is_held(&mddev->reconfig_mutex));
7c7546cc 229 mddev->raid_disks++;
03a9e24e 230 WARN_ONCE(mddev->raid_disks != newconf->raid_disks,
231 "copied raid_disks doesn't match mddev->raid_disks");
232 rcu_assign_pointer(mddev->private, newconf);
1f403624 233 md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
f233ea5c 234 set_capacity(mddev->gendisk, mddev->array_sectors);
3be260cc 235 mddev_resume(mddev);
449aad3e 236 revalidate_disk(mddev->gendisk);
03a9e24e 237 kfree_rcu(oldconf, rcu);
7c7546cc 238 return 0;
1da177e4
LT
239}
240
afa0f557 241static void linear_free(struct mddev *mddev, void *priv)
1da177e4 242{
afa0f557 243 struct linear_conf *conf = priv;
af11c397 244
495d3573 245 kfree(conf);
1da177e4
LT
246}
247
cc27b0c7 248static bool linear_make_request(struct mddev *mddev, struct bio *bio)
1da177e4 249{
20d0189b 250 char b[BDEVNAME_SIZE];
a7120771 251 struct dev_info *tmp_dev;
20d0189b 252 sector_t start_sector, end_sector, data_offset;
868f604b 253 sector_t bio_sector = bio->bi_iter.bi_sector;
1da177e4 254
1eff9d32 255 if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
e9c7469b 256 md_flush_request(mddev, bio);
cc27b0c7 257 return true;
e5dcdd80
N
258 }
259
868f604b
N
260 tmp_dev = which_dev(mddev, bio_sector);
261 start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
262 end_sector = tmp_dev->end_sector;
263 data_offset = tmp_dev->rdev->data_offset;
264
265 if (unlikely(bio_sector >= end_sector ||
266 bio_sector < start_sector))
267 goto out_of_bounds;
268
269 if (unlikely(bio_end_sector(bio) > end_sector)) {
270 /* This bio crosses a device boundary, so we have to split it */
271 struct bio *split = bio_split(bio, end_sector - bio_sector,
272 GFP_NOIO, mddev->bio_set);
273 bio_chain(split, bio);
274 generic_make_request(bio);
275 bio = split;
276 }
f1cad2b6 277
74d46992 278 bio_set_dev(bio, tmp_dev->rdev->bdev);
868f604b
N
279 bio->bi_iter.bi_sector = bio->bi_iter.bi_sector -
280 start_sector + data_offset;
281
282 if (unlikely((bio_op(bio) == REQ_OP_DISCARD) &&
74d46992 283 !blk_queue_discard(bio->bi_disk->queue))) {
868f604b
N
284 /* Just ignore it */
285 bio_endio(bio);
286 } else {
287 if (mddev->gendisk)
74d46992 288 trace_block_bio_remap(bio->bi_disk->queue,
868f604b
N
289 bio, disk_devt(mddev->gendisk),
290 bio_sector);
291 mddev_check_writesame(mddev, bio);
e265eb3a 292 mddev_check_write_zeroes(mddev, bio);
868f604b
N
293 generic_make_request(bio);
294 }
cc27b0c7 295 return true;
20d0189b
KO
296
297out_of_bounds:
a2e202af 298 pr_err("md/linear:%s: make_request: Sector %llu out of bounds on dev %s: %llu sectors, offset %llu\n",
20d0189b
KO
299 mdname(mddev),
300 (unsigned long long)bio->bi_iter.bi_sector,
301 bdevname(tmp_dev->rdev->bdev, b),
302 (unsigned long long)tmp_dev->rdev->sectors,
303 (unsigned long long)start_sector);
304 bio_io_error(bio);
cc27b0c7 305 return true;
1da177e4
LT
306}
307
fd01b88c 308static void linear_status (struct seq_file *seq, struct mddev *mddev)
1da177e4 309{
9d8f0363 310 seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2);
1da177e4
LT
311}
312
3be260cc
N
313static void linear_quiesce(struct mddev *mddev, int state)
314{
315}
316
84fc4b56 317static struct md_personality linear_personality =
1da177e4
LT
318{
319 .name = "linear",
2604b703 320 .level = LEVEL_LINEAR,
1da177e4
LT
321 .owner = THIS_MODULE,
322 .make_request = linear_make_request,
323 .run = linear_run,
afa0f557 324 .free = linear_free,
1da177e4 325 .status = linear_status,
7c7546cc 326 .hot_add_disk = linear_add,
80c3a6ce 327 .size = linear_size,
3be260cc 328 .quiesce = linear_quiesce,
5c675f83 329 .congested = linear_congested,
1da177e4
LT
330};
331
332static int __init linear_init (void)
333{
2604b703 334 return register_md_personality (&linear_personality);
1da177e4
LT
335}
336
337static void linear_exit (void)
338{
2604b703 339 unregister_md_personality (&linear_personality);
1da177e4
LT
340}
341
1da177e4
LT
342module_init(linear_init);
343module_exit(linear_exit);
344MODULE_LICENSE("GPL");
0efb9e61 345MODULE_DESCRIPTION("Linear device concatenation personality for MD");
d9d166c2
N
346MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/
347MODULE_ALIAS("md-linear");
2604b703 348MODULE_ALIAS("md-level--1");