]>
Commit | Line | Data |
---|---|---|
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. | |
13 | ||
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 | |
16 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
17 | */ | |
18 | ||
bff61975 N |
19 | #include <linux/blkdev.h> |
20 | #include <linux/raid/md_u.h> | |
bff61975 | 21 | #include <linux/seq_file.h> |
5a0e3ad6 | 22 | #include <linux/slab.h> |
43b2e5d8 | 23 | #include "md.h" |
ef740c37 | 24 | #include "linear.h" |
1da177e4 | 25 | |
1da177e4 LT |
26 | /* |
27 | * find which device holds a particular offset | |
28 | */ | |
29 | static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector) | |
30 | { | |
aece3d1f | 31 | int lo, mid, hi; |
af11c397 | 32 | linear_conf_t *conf; |
1da177e4 | 33 | |
aece3d1f SS |
34 | lo = 0; |
35 | hi = mddev->raid_disks - 1; | |
af11c397 | 36 | conf = rcu_dereference(mddev->private); |
1da177e4 | 37 | |
aece3d1f SS |
38 | /* |
39 | * Binary Search | |
40 | */ | |
41 | ||
42 | while (hi > lo) { | |
43 | ||
44 | mid = (hi + lo) / 2; | |
45 | if (sector < conf->disks[mid].end_sector) | |
46 | hi = mid; | |
47 | else | |
48 | lo = mid + 1; | |
49 | } | |
50 | ||
51 | return conf->disks + lo; | |
1da177e4 LT |
52 | } |
53 | ||
54 | /** | |
15945fee | 55 | * linear_mergeable_bvec -- tell bio layer if two requests can be merged |
1da177e4 | 56 | * @q: request queue |
cc371e66 | 57 | * @bvm: properties of new bio |
1da177e4 LT |
58 | * @biovec: the request that could be merged to it. |
59 | * | |
60 | * Return amount of bytes we can take at this offset | |
61 | */ | |
cc371e66 AK |
62 | static int linear_mergeable_bvec(struct request_queue *q, |
63 | struct bvec_merge_data *bvm, | |
64 | struct bio_vec *biovec) | |
1da177e4 LT |
65 | { |
66 | mddev_t *mddev = q->queuedata; | |
67 | dev_info_t *dev0; | |
cc371e66 AK |
68 | unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9; |
69 | sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); | |
1da177e4 | 70 | |
af11c397 | 71 | rcu_read_lock(); |
1da177e4 | 72 | dev0 = which_dev(mddev, sector); |
4db7cdc8 | 73 | maxsectors = dev0->end_sector - sector; |
af11c397 | 74 | rcu_read_unlock(); |
1da177e4 LT |
75 | |
76 | if (maxsectors < bio_sectors) | |
77 | maxsectors = 0; | |
78 | else | |
79 | maxsectors -= bio_sectors; | |
80 | ||
81 | if (maxsectors <= (PAGE_SIZE >> 9 ) && bio_sectors == 0) | |
82 | return biovec->bv_len; | |
83 | /* The bytes available at this offset could be really big, | |
84 | * so we cap at 2^31 to avoid overflow */ | |
85 | if (maxsectors > (1 << (31-9))) | |
86 | return 1<<31; | |
87 | return maxsectors << 9; | |
88 | } | |
89 | ||
165125e1 | 90 | static void linear_unplug(struct request_queue *q) |
1da177e4 LT |
91 | { |
92 | mddev_t *mddev = q->queuedata; | |
af11c397 | 93 | linear_conf_t *conf; |
1da177e4 LT |
94 | int i; |
95 | ||
af11c397 S |
96 | rcu_read_lock(); |
97 | conf = rcu_dereference(mddev->private); | |
98 | ||
1da177e4 | 99 | for (i=0; i < mddev->raid_disks; i++) { |
165125e1 | 100 | struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev); |
2ad8b1ef | 101 | blk_unplug(r_queue); |
1da177e4 | 102 | } |
af11c397 | 103 | rcu_read_unlock(); |
1da177e4 LT |
104 | } |
105 | ||
26be34dc N |
106 | static int linear_congested(void *data, int bits) |
107 | { | |
108 | mddev_t *mddev = data; | |
af11c397 | 109 | linear_conf_t *conf; |
26be34dc N |
110 | int i, ret = 0; |
111 | ||
3fa841d7 N |
112 | if (mddev_congested(mddev, bits)) |
113 | return 1; | |
114 | ||
af11c397 S |
115 | rcu_read_lock(); |
116 | conf = rcu_dereference(mddev->private); | |
117 | ||
26be34dc | 118 | for (i = 0; i < mddev->raid_disks && !ret ; i++) { |
165125e1 | 119 | struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev); |
26be34dc N |
120 | ret |= bdi_congested(&q->backing_dev_info, bits); |
121 | } | |
af11c397 S |
122 | |
123 | rcu_read_unlock(); | |
26be34dc N |
124 | return ret; |
125 | } | |
126 | ||
80c3a6ce DW |
127 | static sector_t linear_size(mddev_t *mddev, sector_t sectors, int raid_disks) |
128 | { | |
af11c397 S |
129 | linear_conf_t *conf; |
130 | sector_t array_sectors; | |
80c3a6ce | 131 | |
af11c397 S |
132 | rcu_read_lock(); |
133 | conf = rcu_dereference(mddev->private); | |
80c3a6ce DW |
134 | WARN_ONCE(sectors || raid_disks, |
135 | "%s does not support generic reshape\n", __func__); | |
af11c397 S |
136 | array_sectors = conf->array_sectors; |
137 | rcu_read_unlock(); | |
80c3a6ce | 138 | |
af11c397 | 139 | return array_sectors; |
80c3a6ce DW |
140 | } |
141 | ||
7c7546cc | 142 | static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks) |
1da177e4 LT |
143 | { |
144 | linear_conf_t *conf; | |
1da177e4 | 145 | mdk_rdev_t *rdev; |
45d4582f | 146 | int i, cnt; |
1da177e4 | 147 | |
7c7546cc | 148 | conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t), |
1da177e4 LT |
149 | GFP_KERNEL); |
150 | if (!conf) | |
7c7546cc N |
151 | return NULL; |
152 | ||
1da177e4 | 153 | cnt = 0; |
d6e22150 | 154 | conf->array_sectors = 0; |
1da177e4 | 155 | |
159ec1fc | 156 | list_for_each_entry(rdev, &mddev->disks, same_set) { |
1da177e4 LT |
157 | int j = rdev->raid_disk; |
158 | dev_info_t *disk = conf->disks + j; | |
13f2682b | 159 | sector_t sectors; |
1da177e4 | 160 | |
13864515 | 161 | if (j < 0 || j >= raid_disks || disk->rdev) { |
2dc40f80 N |
162 | printk(KERN_ERR "md/linear:%s: disk numbering problem. Aborting!\n", |
163 | mdname(mddev)); | |
1da177e4 LT |
164 | goto out; |
165 | } | |
166 | ||
167 | disk->rdev = rdev; | |
13f2682b N |
168 | if (mddev->chunk_sectors) { |
169 | sectors = rdev->sectors; | |
170 | sector_div(sectors, mddev->chunk_sectors); | |
171 | rdev->sectors = sectors * mddev->chunk_sectors; | |
172 | } | |
1da177e4 | 173 | |
8f6c2e4b MP |
174 | disk_stack_limits(mddev->gendisk, rdev->bdev, |
175 | rdev->data_offset << 9); | |
1da177e4 | 176 | /* as we don't honour merge_bvec_fn, we must never risk |
627a2d3c N |
177 | * violating it, so limit max_segments to 1 lying within |
178 | * a single page. | |
1da177e4 | 179 | */ |
627a2d3c N |
180 | if (rdev->bdev->bd_disk->queue->merge_bvec_fn) { |
181 | blk_queue_max_segments(mddev->queue, 1); | |
182 | blk_queue_segment_boundary(mddev->queue, | |
183 | PAGE_CACHE_SIZE - 1); | |
184 | } | |
1da177e4 | 185 | |
dd8ac336 | 186 | conf->array_sectors += rdev->sectors; |
1da177e4 | 187 | cnt++; |
4db7cdc8 | 188 | |
1da177e4 | 189 | } |
7c7546cc | 190 | if (cnt != raid_disks) { |
2dc40f80 N |
191 | printk(KERN_ERR "md/linear:%s: not enough drives present. Aborting!\n", |
192 | mdname(mddev)); | |
1da177e4 LT |
193 | goto out; |
194 | } | |
195 | ||
196 | /* | |
45d4582f | 197 | * Here we calculate the device offsets. |
1da177e4 | 198 | */ |
4db7cdc8 SS |
199 | conf->disks[0].end_sector = conf->disks[0].rdev->sectors; |
200 | ||
a778b73f | 201 | for (i = 1; i < raid_disks; i++) |
4db7cdc8 SS |
202 | conf->disks[i].end_sector = |
203 | conf->disks[i-1].end_sector + | |
204 | conf->disks[i].rdev->sectors; | |
15945fee | 205 | |
7c7546cc N |
206 | return conf; |
207 | ||
208 | out: | |
209 | kfree(conf); | |
210 | return NULL; | |
211 | } | |
212 | ||
213 | static int linear_run (mddev_t *mddev) | |
214 | { | |
215 | linear_conf_t *conf; | |
216 | ||
0894cc30 AN |
217 | if (md_check_no_bitmap(mddev)) |
218 | return -EINVAL; | |
e7e72bf6 | 219 | mddev->queue->queue_lock = &mddev->queue->__queue_lock; |
7c7546cc N |
220 | conf = linear_conf(mddev, mddev->raid_disks); |
221 | ||
222 | if (!conf) | |
223 | return 1; | |
224 | mddev->private = conf; | |
1f403624 | 225 | md_set_array_sectors(mddev, linear_size(mddev, 0, 0)); |
7c7546cc | 226 | |
1da177e4 LT |
227 | blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec); |
228 | mddev->queue->unplug_fn = linear_unplug; | |
26be34dc N |
229 | mddev->queue->backing_dev_info.congested_fn = linear_congested; |
230 | mddev->queue->backing_dev_info.congested_data = mddev; | |
ac5e7113 | 231 | md_integrity_register(mddev); |
1da177e4 | 232 | return 0; |
7c7546cc | 233 | } |
1da177e4 | 234 | |
495d3573 N |
235 | static void free_conf(struct rcu_head *head) |
236 | { | |
237 | linear_conf_t *conf = container_of(head, linear_conf_t, rcu); | |
238 | kfree(conf); | |
239 | } | |
240 | ||
7c7546cc N |
241 | static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev) |
242 | { | |
243 | /* Adding a drive to a linear array allows the array to grow. | |
244 | * It is permitted if the new drive has a matching superblock | |
245 | * already on it, with raid_disk equal to raid_disks. | |
246 | * It is achieved by creating a new linear_private_data structure | |
247 | * and swapping it in in-place of the current one. | |
248 | * The current one is never freed until the array is stopped. | |
249 | * This avoids races. | |
250 | */ | |
495d3573 | 251 | linear_conf_t *newconf, *oldconf; |
7c7546cc | 252 | |
a778b73f | 253 | if (rdev->saved_raid_disk != mddev->raid_disks) |
7c7546cc N |
254 | return -EINVAL; |
255 | ||
a778b73f N |
256 | rdev->raid_disk = rdev->saved_raid_disk; |
257 | ||
7c7546cc N |
258 | newconf = linear_conf(mddev,mddev->raid_disks+1); |
259 | ||
260 | if (!newconf) | |
261 | return -ENOMEM; | |
262 | ||
495d3573 | 263 | oldconf = rcu_dereference(mddev->private); |
7c7546cc | 264 | mddev->raid_disks++; |
af11c397 | 265 | rcu_assign_pointer(mddev->private, newconf); |
1f403624 | 266 | md_set_array_sectors(mddev, linear_size(mddev, 0, 0)); |
f233ea5c | 267 | set_capacity(mddev->gendisk, mddev->array_sectors); |
449aad3e | 268 | revalidate_disk(mddev->gendisk); |
495d3573 | 269 | call_rcu(&oldconf->rcu, free_conf); |
7c7546cc | 270 | return 0; |
1da177e4 LT |
271 | } |
272 | ||
273 | static int linear_stop (mddev_t *mddev) | |
274 | { | |
070ec55d | 275 | linear_conf_t *conf = mddev->private; |
af11c397 S |
276 | |
277 | /* | |
278 | * We do not require rcu protection here since | |
279 | * we hold reconfig_mutex for both linear_add and | |
280 | * linear_stop, so they cannot race. | |
495d3573 N |
281 | * We should make sure any old 'conf's are properly |
282 | * freed though. | |
af11c397 | 283 | */ |
495d3573 | 284 | rcu_barrier(); |
1da177e4 | 285 | blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ |
495d3573 | 286 | kfree(conf); |
ef2f80ff | 287 | mddev->private = NULL; |
1da177e4 LT |
288 | |
289 | return 0; | |
290 | } | |
291 | ||
21a52c6d | 292 | static int linear_make_request (mddev_t *mddev, struct bio *bio) |
1da177e4 | 293 | { |
1da177e4 | 294 | dev_info_t *tmp_dev; |
4db7cdc8 | 295 | sector_t start_sector; |
1da177e4 | 296 | |
e9c7469b TH |
297 | if (unlikely(bio->bi_rw & REQ_FLUSH)) { |
298 | md_flush_request(mddev, bio); | |
e5dcdd80 N |
299 | return 0; |
300 | } | |
301 | ||
af11c397 | 302 | rcu_read_lock(); |
1da177e4 | 303 | tmp_dev = which_dev(mddev, bio->bi_sector); |
4db7cdc8 SS |
304 | start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors; |
305 | ||
af11c397 | 306 | |
4db7cdc8 SS |
307 | if (unlikely(bio->bi_sector >= (tmp_dev->end_sector) |
308 | || (bio->bi_sector < start_sector))) { | |
1da177e4 LT |
309 | char b[BDEVNAME_SIZE]; |
310 | ||
2dc40f80 N |
311 | printk(KERN_ERR |
312 | "md/linear:%s: make_request: Sector %llu out of bounds on " | |
313 | "dev %s: %llu sectors, offset %llu\n", | |
314 | mdname(mddev), | |
315 | (unsigned long long)bio->bi_sector, | |
316 | bdevname(tmp_dev->rdev->bdev, b), | |
317 | (unsigned long long)tmp_dev->rdev->sectors, | |
318 | (unsigned long long)start_sector); | |
af11c397 | 319 | rcu_read_unlock(); |
6712ecf8 | 320 | bio_io_error(bio); |
1da177e4 LT |
321 | return 0; |
322 | } | |
323 | if (unlikely(bio->bi_sector + (bio->bi_size >> 9) > | |
4db7cdc8 | 324 | tmp_dev->end_sector)) { |
1da177e4 LT |
325 | /* This bio crosses a device boundary, so we have to |
326 | * split it. | |
327 | */ | |
328 | struct bio_pair *bp; | |
af11c397 S |
329 | sector_t end_sector = tmp_dev->end_sector; |
330 | ||
331 | rcu_read_unlock(); | |
6283815d | 332 | |
af11c397 | 333 | bp = bio_split(bio, end_sector - bio->bi_sector); |
6283815d | 334 | |
21a52c6d | 335 | if (linear_make_request(mddev, &bp->bio1)) |
1da177e4 | 336 | generic_make_request(&bp->bio1); |
21a52c6d | 337 | if (linear_make_request(mddev, &bp->bio2)) |
1da177e4 LT |
338 | generic_make_request(&bp->bio2); |
339 | bio_pair_release(bp); | |
340 | return 0; | |
341 | } | |
342 | ||
343 | bio->bi_bdev = tmp_dev->rdev->bdev; | |
4db7cdc8 | 344 | bio->bi_sector = bio->bi_sector - start_sector |
6283815d | 345 | + tmp_dev->rdev->data_offset; |
af11c397 | 346 | rcu_read_unlock(); |
1da177e4 LT |
347 | |
348 | return 1; | |
349 | } | |
350 | ||
351 | static void linear_status (struct seq_file *seq, mddev_t *mddev) | |
352 | { | |
353 | ||
9d8f0363 | 354 | seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2); |
1da177e4 LT |
355 | } |
356 | ||
357 | ||
2604b703 | 358 | static struct mdk_personality linear_personality = |
1da177e4 LT |
359 | { |
360 | .name = "linear", | |
2604b703 | 361 | .level = LEVEL_LINEAR, |
1da177e4 LT |
362 | .owner = THIS_MODULE, |
363 | .make_request = linear_make_request, | |
364 | .run = linear_run, | |
365 | .stop = linear_stop, | |
366 | .status = linear_status, | |
7c7546cc | 367 | .hot_add_disk = linear_add, |
80c3a6ce | 368 | .size = linear_size, |
1da177e4 LT |
369 | }; |
370 | ||
371 | static int __init linear_init (void) | |
372 | { | |
2604b703 | 373 | return register_md_personality (&linear_personality); |
1da177e4 LT |
374 | } |
375 | ||
376 | static void linear_exit (void) | |
377 | { | |
2604b703 | 378 | unregister_md_personality (&linear_personality); |
1da177e4 LT |
379 | } |
380 | ||
381 | ||
382 | module_init(linear_init); | |
383 | module_exit(linear_exit); | |
384 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 385 | MODULE_DESCRIPTION("Linear device concatenation personality for MD"); |
d9d166c2 N |
386 | MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/ |
387 | MODULE_ALIAS("md-linear"); | |
2604b703 | 388 | MODULE_ALIAS("md-level--1"); |