Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | raid0.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 | Copyright (C) 1999, 2000 Ingo Molnar, Red Hat | |
7 | ||
8 | ||
9 | RAID-0 management functions. | |
10 | ||
11 | This program is free software; you can redistribute it and/or modify | |
12 | it under the terms of the GNU General Public License as published by | |
13 | the Free Software Foundation; either version 2, or (at your option) | |
14 | any later version. | |
15 | ||
16 | You should have received a copy of the GNU General Public License | |
17 | (for example /usr/src/linux/COPYING); if not, write to the Free | |
18 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
19 | */ | |
20 | ||
bff61975 | 21 | #include <linux/blkdev.h> |
bff61975 | 22 | #include <linux/seq_file.h> |
5a0e3ad6 | 23 | #include <linux/slab.h> |
43b2e5d8 | 24 | #include "md.h" |
ef740c37 | 25 | #include "raid0.h" |
9af204cf | 26 | #include "raid5.h" |
1da177e4 | 27 | |
26be34dc N |
28 | static int raid0_congested(void *data, int bits) |
29 | { | |
30 | mddev_t *mddev = data; | |
070ec55d | 31 | raid0_conf_t *conf = mddev->private; |
b414579f | 32 | mdk_rdev_t **devlist = conf->devlist; |
84707f38 | 33 | int raid_disks = conf->strip_zone[0].nb_dev; |
26be34dc N |
34 | int i, ret = 0; |
35 | ||
3fa841d7 N |
36 | if (mddev_congested(mddev, bits)) |
37 | return 1; | |
38 | ||
84707f38 | 39 | for (i = 0; i < raid_disks && !ret ; i++) { |
165125e1 | 40 | struct request_queue *q = bdev_get_queue(devlist[i]->bdev); |
26be34dc N |
41 | |
42 | ret |= bdi_congested(&q->backing_dev_info, bits); | |
43 | } | |
44 | return ret; | |
45 | } | |
46 | ||
46994191 | 47 | /* |
48 | * inform the user of the raid configuration | |
49 | */ | |
50 | static void dump_zones(mddev_t *mddev) | |
51 | { | |
52 | int j, k, h; | |
53 | sector_t zone_size = 0; | |
54 | sector_t zone_start = 0; | |
55 | char b[BDEVNAME_SIZE]; | |
56 | raid0_conf_t *conf = mddev->private; | |
84707f38 | 57 | int raid_disks = conf->strip_zone[0].nb_dev; |
46994191 | 58 | printk(KERN_INFO "******* %s configuration *********\n", |
59 | mdname(mddev)); | |
60 | h = 0; | |
61 | for (j = 0; j < conf->nr_strip_zones; j++) { | |
62 | printk(KERN_INFO "zone%d=[", j); | |
63 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
b5a20961 | 64 | printk(KERN_CONT "%s/", |
84707f38 | 65 | bdevname(conf->devlist[j*raid_disks |
46994191 | 66 | + k]->bdev, b)); |
b5a20961 | 67 | printk(KERN_CONT "]\n"); |
46994191 | 68 | |
69 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
70 | printk(KERN_INFO " zone offset=%llukb " | |
71 | "device offset=%llukb size=%llukb\n", | |
72 | (unsigned long long)zone_start>>1, | |
73 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
74 | (unsigned long long)zone_size>>1); | |
75 | zone_start = conf->strip_zone[j].zone_end; | |
76 | } | |
77 | printk(KERN_INFO "**********************************\n\n"); | |
78 | } | |
79 | ||
9af204cf | 80 | static int create_strip_zones(mddev_t *mddev, raid0_conf_t **private_conf) |
1da177e4 | 81 | { |
a9f326eb | 82 | int i, c, err; |
49f357a2 | 83 | sector_t curr_zone_end, sectors; |
b414579f | 84 | mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev; |
1da177e4 LT |
85 | struct strip_zone *zone; |
86 | int cnt; | |
87 | char b[BDEVNAME_SIZE]; | |
ed7b0038 AN |
88 | raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL); |
89 | ||
90 | if (!conf) | |
91 | return -ENOMEM; | |
159ec1fc | 92 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
b5a20961 N |
93 | printk(KERN_INFO "md/raid0:%s: looking at %s\n", |
94 | mdname(mddev), | |
95 | bdevname(rdev1->bdev, b)); | |
1da177e4 | 96 | c = 0; |
13f2682b N |
97 | |
98 | /* round size to chunk_size */ | |
99 | sectors = rdev1->sectors; | |
100 | sector_div(sectors, mddev->chunk_sectors); | |
101 | rdev1->sectors = sectors * mddev->chunk_sectors; | |
102 | ||
159ec1fc | 103 | list_for_each_entry(rdev2, &mddev->disks, same_set) { |
b5a20961 N |
104 | printk(KERN_INFO "md/raid0:%s: comparing %s(%llu)", |
105 | mdname(mddev), | |
1da177e4 | 106 | bdevname(rdev1->bdev,b), |
dd8ac336 | 107 | (unsigned long long)rdev1->sectors); |
b5a20961 | 108 | printk(KERN_CONT " with %s(%llu)\n", |
1da177e4 | 109 | bdevname(rdev2->bdev,b), |
dd8ac336 | 110 | (unsigned long long)rdev2->sectors); |
1da177e4 | 111 | if (rdev2 == rdev1) { |
b5a20961 N |
112 | printk(KERN_INFO "md/raid0:%s: END\n", |
113 | mdname(mddev)); | |
1da177e4 LT |
114 | break; |
115 | } | |
dd8ac336 | 116 | if (rdev2->sectors == rdev1->sectors) { |
1da177e4 LT |
117 | /* |
118 | * Not unique, don't count it as a new | |
119 | * group | |
120 | */ | |
b5a20961 N |
121 | printk(KERN_INFO "md/raid0:%s: EQUAL\n", |
122 | mdname(mddev)); | |
1da177e4 LT |
123 | c = 1; |
124 | break; | |
125 | } | |
b5a20961 N |
126 | printk(KERN_INFO "md/raid0:%s: NOT EQUAL\n", |
127 | mdname(mddev)); | |
1da177e4 LT |
128 | } |
129 | if (!c) { | |
b5a20961 N |
130 | printk(KERN_INFO "md/raid0:%s: ==> UNIQUE\n", |
131 | mdname(mddev)); | |
1da177e4 | 132 | conf->nr_strip_zones++; |
b5a20961 N |
133 | printk(KERN_INFO "md/raid0:%s: %d zones\n", |
134 | mdname(mddev), conf->nr_strip_zones); | |
1da177e4 LT |
135 | } |
136 | } | |
b5a20961 N |
137 | printk(KERN_INFO "md/raid0:%s: FINAL %d zones\n", |
138 | mdname(mddev), conf->nr_strip_zones); | |
ed7b0038 | 139 | err = -ENOMEM; |
9ffae0cf | 140 | conf->strip_zone = kzalloc(sizeof(struct strip_zone)* |
1da177e4 LT |
141 | conf->nr_strip_zones, GFP_KERNEL); |
142 | if (!conf->strip_zone) | |
ed7b0038 | 143 | goto abort; |
9ffae0cf | 144 | conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* |
1da177e4 LT |
145 | conf->nr_strip_zones*mddev->raid_disks, |
146 | GFP_KERNEL); | |
147 | if (!conf->devlist) | |
ed7b0038 | 148 | goto abort; |
1da177e4 | 149 | |
1da177e4 LT |
150 | /* The first zone must contain all devices, so here we check that |
151 | * there is a proper alignment of slots to devices and find them all | |
152 | */ | |
153 | zone = &conf->strip_zone[0]; | |
154 | cnt = 0; | |
155 | smallest = NULL; | |
b414579f | 156 | dev = conf->devlist; |
ed7b0038 | 157 | err = -EINVAL; |
159ec1fc | 158 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
1da177e4 LT |
159 | int j = rdev1->raid_disk; |
160 | ||
e93f68a1 | 161 | if (mddev->level == 10) { |
9af204cf TM |
162 | /* taking over a raid10-n2 array */ |
163 | j /= 2; | |
e93f68a1 N |
164 | rdev1->new_raid_disk = j; |
165 | } | |
9af204cf | 166 | |
fc3a08b8 KW |
167 | if (mddev->level == 1) { |
168 | /* taiking over a raid1 array- | |
169 | * we have only one active disk | |
170 | */ | |
171 | j = 0; | |
172 | rdev1->new_raid_disk = j; | |
173 | } | |
174 | ||
1da177e4 | 175 | if (j < 0 || j >= mddev->raid_disks) { |
b5a20961 N |
176 | printk(KERN_ERR "md/raid0:%s: bad disk number %d - " |
177 | "aborting!\n", mdname(mddev), j); | |
1da177e4 LT |
178 | goto abort; |
179 | } | |
b414579f | 180 | if (dev[j]) { |
b5a20961 N |
181 | printk(KERN_ERR "md/raid0:%s: multiple devices for %d - " |
182 | "aborting!\n", mdname(mddev), j); | |
1da177e4 LT |
183 | goto abort; |
184 | } | |
b414579f | 185 | dev[j] = rdev1; |
1da177e4 | 186 | |
8f6c2e4b MP |
187 | disk_stack_limits(mddev->gendisk, rdev1->bdev, |
188 | rdev1->data_offset << 9); | |
1da177e4 | 189 | /* as we don't honour merge_bvec_fn, we must never risk |
627a2d3c N |
190 | * violating it, so limit ->max_segments to 1, lying within |
191 | * a single page. | |
1da177e4 LT |
192 | */ |
193 | ||
627a2d3c N |
194 | if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) { |
195 | blk_queue_max_segments(mddev->queue, 1); | |
196 | blk_queue_segment_boundary(mddev->queue, | |
197 | PAGE_CACHE_SIZE - 1); | |
198 | } | |
dd8ac336 | 199 | if (!smallest || (rdev1->sectors < smallest->sectors)) |
1da177e4 LT |
200 | smallest = rdev1; |
201 | cnt++; | |
202 | } | |
203 | if (cnt != mddev->raid_disks) { | |
b5a20961 N |
204 | printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - " |
205 | "aborting!\n", mdname(mddev), cnt, mddev->raid_disks); | |
1da177e4 LT |
206 | goto abort; |
207 | } | |
208 | zone->nb_dev = cnt; | |
49f357a2 | 209 | zone->zone_end = smallest->sectors * cnt; |
1da177e4 | 210 | |
49f357a2 | 211 | curr_zone_end = zone->zone_end; |
1da177e4 LT |
212 | |
213 | /* now do the other zones */ | |
214 | for (i = 1; i < conf->nr_strip_zones; i++) | |
215 | { | |
a9f326eb N |
216 | int j; |
217 | ||
1da177e4 | 218 | zone = conf->strip_zone + i; |
b414579f | 219 | dev = conf->devlist + i * mddev->raid_disks; |
1da177e4 | 220 | |
b5a20961 N |
221 | printk(KERN_INFO "md/raid0:%s: zone %d\n", |
222 | mdname(mddev), i); | |
d27a43ab | 223 | zone->dev_start = smallest->sectors; |
1da177e4 LT |
224 | smallest = NULL; |
225 | c = 0; | |
226 | ||
227 | for (j=0; j<cnt; j++) { | |
b414579f | 228 | rdev = conf->devlist[j]; |
b5a20961 N |
229 | printk(KERN_INFO "md/raid0:%s: checking %s ...", |
230 | mdname(mddev), | |
231 | bdevname(rdev->bdev, b)); | |
d27a43ab | 232 | if (rdev->sectors <= zone->dev_start) { |
b5a20961 | 233 | printk(KERN_CONT " nope.\n"); |
dd8ac336 AN |
234 | continue; |
235 | } | |
b5a20961 | 236 | printk(KERN_CONT " contained as device %d\n", c); |
b414579f | 237 | dev[c] = rdev; |
dd8ac336 AN |
238 | c++; |
239 | if (!smallest || rdev->sectors < smallest->sectors) { | |
240 | smallest = rdev; | |
b5a20961 N |
241 | printk(KERN_INFO "md/raid0:%s: (%llu) is smallest!.\n", |
242 | mdname(mddev), | |
243 | (unsigned long long)rdev->sectors); | |
dd8ac336 | 244 | } |
1da177e4 LT |
245 | } |
246 | ||
247 | zone->nb_dev = c; | |
49f357a2 | 248 | sectors = (smallest->sectors - zone->dev_start) * c; |
b5a20961 N |
249 | printk(KERN_INFO "md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n", |
250 | mdname(mddev), | |
251 | zone->nb_dev, (unsigned long long)sectors); | |
1da177e4 | 252 | |
49f357a2 | 253 | curr_zone_end += sectors; |
d27a43ab | 254 | zone->zone_end = curr_zone_end; |
1da177e4 | 255 | |
b5a20961 N |
256 | printk(KERN_INFO "md/raid0:%s: current zone start: %llu\n", |
257 | mdname(mddev), | |
258 | (unsigned long long)smallest->sectors); | |
1da177e4 | 259 | } |
26be34dc N |
260 | mddev->queue->backing_dev_info.congested_fn = raid0_congested; |
261 | mddev->queue->backing_dev_info.congested_data = mddev; | |
1da177e4 | 262 | |
92e59b6b | 263 | /* |
264 | * now since we have the hard sector sizes, we can make sure | |
265 | * chunk size is a multiple of that sector size | |
266 | */ | |
9d8f0363 | 267 | if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) { |
b5a20961 | 268 | printk(KERN_ERR "md/raid0:%s: chunk_size of %d not valid\n", |
92e59b6b | 269 | mdname(mddev), |
9d8f0363 | 270 | mddev->chunk_sectors << 9); |
92e59b6b | 271 | goto abort; |
272 | } | |
8f6c2e4b MP |
273 | |
274 | blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); | |
275 | blk_queue_io_opt(mddev->queue, | |
276 | (mddev->chunk_sectors << 9) * mddev->raid_disks); | |
277 | ||
b5a20961 | 278 | printk(KERN_INFO "md/raid0:%s: done.\n", mdname(mddev)); |
9af204cf TM |
279 | *private_conf = conf; |
280 | ||
1da177e4 | 281 | return 0; |
5568a603 | 282 | abort: |
ed7b0038 AN |
283 | kfree(conf->strip_zone); |
284 | kfree(conf->devlist); | |
285 | kfree(conf); | |
9af204cf | 286 | *private_conf = NULL; |
ed7b0038 | 287 | return err; |
1da177e4 LT |
288 | } |
289 | ||
290 | /** | |
291 | * raid0_mergeable_bvec -- tell bio layer if a two requests can be merged | |
292 | * @q: request queue | |
cc371e66 | 293 | * @bvm: properties of new bio |
1da177e4 LT |
294 | * @biovec: the request that could be merged to it. |
295 | * | |
296 | * Return amount of bytes we can accept at this offset | |
297 | */ | |
cc371e66 AK |
298 | static int raid0_mergeable_bvec(struct request_queue *q, |
299 | struct bvec_merge_data *bvm, | |
300 | struct bio_vec *biovec) | |
1da177e4 LT |
301 | { |
302 | mddev_t *mddev = q->queuedata; | |
cc371e66 | 303 | sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); |
1da177e4 | 304 | int max; |
9d8f0363 | 305 | unsigned int chunk_sectors = mddev->chunk_sectors; |
cc371e66 | 306 | unsigned int bio_sectors = bvm->bi_size >> 9; |
1da177e4 | 307 | |
d6e412ea | 308 | if (is_power_of_2(chunk_sectors)) |
fbb704ef | 309 | max = (chunk_sectors - ((sector & (chunk_sectors-1)) |
310 | + bio_sectors)) << 9; | |
311 | else | |
312 | max = (chunk_sectors - (sector_div(sector, chunk_sectors) | |
313 | + bio_sectors)) << 9; | |
1da177e4 LT |
314 | if (max < 0) max = 0; /* bio_add cannot handle a negative return */ |
315 | if (max <= biovec->bv_len && bio_sectors == 0) | |
316 | return biovec->bv_len; | |
317 | else | |
318 | return max; | |
319 | } | |
320 | ||
80c3a6ce DW |
321 | static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks) |
322 | { | |
323 | sector_t array_sectors = 0; | |
324 | mdk_rdev_t *rdev; | |
325 | ||
326 | WARN_ONCE(sectors || raid_disks, | |
327 | "%s does not support generic reshape\n", __func__); | |
328 | ||
329 | list_for_each_entry(rdev, &mddev->disks, same_set) | |
330 | array_sectors += rdev->sectors; | |
331 | ||
332 | return array_sectors; | |
333 | } | |
334 | ||
8f79cfcd | 335 | static int raid0_run(mddev_t *mddev) |
1da177e4 | 336 | { |
9af204cf | 337 | raid0_conf_t *conf; |
5568a603 | 338 | int ret; |
1da177e4 | 339 | |
9d8f0363 | 340 | if (mddev->chunk_sectors == 0) { |
b5a20961 N |
341 | printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n", |
342 | mdname(mddev)); | |
2604b703 N |
343 | return -EINVAL; |
344 | } | |
0894cc30 AN |
345 | if (md_check_no_bitmap(mddev)) |
346 | return -EINVAL; | |
086fa5ff | 347 | blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); |
1da177e4 | 348 | |
9af204cf TM |
349 | /* if private is not null, we are here after takeover */ |
350 | if (mddev->private == NULL) { | |
351 | ret = create_strip_zones(mddev, &conf); | |
352 | if (ret < 0) | |
353 | return ret; | |
354 | mddev->private = conf; | |
355 | } | |
356 | conf = mddev->private; | |
1da177e4 LT |
357 | |
358 | /* calculate array device size */ | |
1f403624 | 359 | md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); |
1da177e4 | 360 | |
b5a20961 N |
361 | printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n", |
362 | mdname(mddev), | |
363 | (unsigned long long)mddev->array_sectors); | |
1da177e4 LT |
364 | /* calculate the max read-ahead size. |
365 | * For read-ahead of large files to be effective, we need to | |
366 | * readahead at least twice a whole stripe. i.e. number of devices | |
367 | * multiplied by chunk size times 2. | |
368 | * If an individual device has an ra_pages greater than the | |
369 | * chunk size, then we will not drive that device as hard as it | |
370 | * wants. We consider this a configuration error: a larger | |
371 | * chunksize should be used in that case. | |
372 | */ | |
373 | { | |
9d8f0363 AN |
374 | int stripe = mddev->raid_disks * |
375 | (mddev->chunk_sectors << 9) / PAGE_SIZE; | |
1da177e4 LT |
376 | if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) |
377 | mddev->queue->backing_dev_info.ra_pages = 2* stripe; | |
378 | } | |
379 | ||
1da177e4 | 380 | blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); |
46994191 | 381 | dump_zones(mddev); |
a91a2785 | 382 | return md_integrity_register(mddev); |
1da177e4 LT |
383 | } |
384 | ||
fb5ab4b5 | 385 | static int raid0_stop(mddev_t *mddev) |
1da177e4 | 386 | { |
070ec55d | 387 | raid0_conf_t *conf = mddev->private; |
1da177e4 LT |
388 | |
389 | blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ | |
990a8baf | 390 | kfree(conf->strip_zone); |
fb5ab4b5 | 391 | kfree(conf->devlist); |
990a8baf | 392 | kfree(conf); |
1da177e4 | 393 | mddev->private = NULL; |
1da177e4 LT |
394 | return 0; |
395 | } | |
396 | ||
49f357a2 N |
397 | /* Find the zone which holds a particular offset |
398 | * Update *sectorp to be an offset in that zone | |
399 | */ | |
dc582663 | 400 | static struct strip_zone *find_zone(struct raid0_private_data *conf, |
49f357a2 | 401 | sector_t *sectorp) |
dc582663 AN |
402 | { |
403 | int i; | |
404 | struct strip_zone *z = conf->strip_zone; | |
49f357a2 | 405 | sector_t sector = *sectorp; |
dc582663 AN |
406 | |
407 | for (i = 0; i < conf->nr_strip_zones; i++) | |
49f357a2 N |
408 | if (sector < z[i].zone_end) { |
409 | if (i) | |
410 | *sectorp = sector - z[i-1].zone_end; | |
dc582663 | 411 | return z + i; |
49f357a2 | 412 | } |
dc582663 AN |
413 | BUG(); |
414 | } | |
415 | ||
fbb704ef | 416 | /* |
417 | * remaps the bio to the target device. we separate two flows. | |
418 | * power 2 flow and a general flow for the sake of perfromance | |
419 | */ | |
420 | static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone, | |
421 | sector_t sector, sector_t *sector_offset) | |
1da177e4 | 422 | { |
fbb704ef | 423 | unsigned int sect_in_chunk; |
424 | sector_t chunk; | |
070ec55d | 425 | raid0_conf_t *conf = mddev->private; |
84707f38 | 426 | int raid_disks = conf->strip_zone[0].nb_dev; |
9d8f0363 | 427 | unsigned int chunk_sects = mddev->chunk_sectors; |
fbb704ef | 428 | |
d6e412ea | 429 | if (is_power_of_2(chunk_sects)) { |
fbb704ef | 430 | int chunksect_bits = ffz(~chunk_sects); |
431 | /* find the sector offset inside the chunk */ | |
432 | sect_in_chunk = sector & (chunk_sects - 1); | |
433 | sector >>= chunksect_bits; | |
434 | /* chunk in zone */ | |
435 | chunk = *sector_offset; | |
436 | /* quotient is the chunk in real device*/ | |
437 | sector_div(chunk, zone->nb_dev << chunksect_bits); | |
438 | } else{ | |
439 | sect_in_chunk = sector_div(sector, chunk_sects); | |
440 | chunk = *sector_offset; | |
441 | sector_div(chunk, chunk_sects * zone->nb_dev); | |
442 | } | |
443 | /* | |
444 | * position the bio over the real device | |
445 | * real sector = chunk in device + starting of zone | |
446 | * + the position in the chunk | |
447 | */ | |
448 | *sector_offset = (chunk * chunk_sects) + sect_in_chunk; | |
84707f38 | 449 | return conf->devlist[(zone - conf->strip_zone)*raid_disks |
fbb704ef | 450 | + sector_div(sector, zone->nb_dev)]; |
451 | } | |
452 | ||
453 | /* | |
454 | * Is io distribute over 1 or more chunks ? | |
455 | */ | |
456 | static inline int is_io_in_chunk_boundary(mddev_t *mddev, | |
457 | unsigned int chunk_sects, struct bio *bio) | |
458 | { | |
d6e412ea | 459 | if (likely(is_power_of_2(chunk_sects))) { |
fbb704ef | 460 | return chunk_sects >= ((bio->bi_sector & (chunk_sects-1)) |
461 | + (bio->bi_size >> 9)); | |
462 | } else{ | |
463 | sector_t sector = bio->bi_sector; | |
464 | return chunk_sects >= (sector_div(sector, chunk_sects) | |
465 | + (bio->bi_size >> 9)); | |
466 | } | |
467 | } | |
468 | ||
21a52c6d | 469 | static int raid0_make_request(mddev_t *mddev, struct bio *bio) |
fbb704ef | 470 | { |
fbb704ef | 471 | unsigned int chunk_sects; |
472 | sector_t sector_offset; | |
1da177e4 LT |
473 | struct strip_zone *zone; |
474 | mdk_rdev_t *tmp_dev; | |
1da177e4 | 475 | |
e9c7469b TH |
476 | if (unlikely(bio->bi_rw & REQ_FLUSH)) { |
477 | md_flush_request(mddev, bio); | |
e5dcdd80 N |
478 | return 0; |
479 | } | |
480 | ||
9d8f0363 | 481 | chunk_sects = mddev->chunk_sectors; |
fbb704ef | 482 | if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) { |
483 | sector_t sector = bio->bi_sector; | |
1da177e4 LT |
484 | struct bio_pair *bp; |
485 | /* Sanity check -- queue functions should prevent this happening */ | |
486 | if (bio->bi_vcnt != 1 || | |
487 | bio->bi_idx != 0) | |
488 | goto bad_map; | |
489 | /* This is a one page bio that upper layers | |
490 | * refuse to split for us, so we need to split it. | |
491 | */ | |
d6e412ea | 492 | if (likely(is_power_of_2(chunk_sects))) |
fbb704ef | 493 | bp = bio_split(bio, chunk_sects - (sector & |
494 | (chunk_sects-1))); | |
495 | else | |
496 | bp = bio_split(bio, chunk_sects - | |
497 | sector_div(sector, chunk_sects)); | |
21a52c6d | 498 | if (raid0_make_request(mddev, &bp->bio1)) |
1da177e4 | 499 | generic_make_request(&bp->bio1); |
21a52c6d | 500 | if (raid0_make_request(mddev, &bp->bio2)) |
1da177e4 LT |
501 | generic_make_request(&bp->bio2); |
502 | ||
503 | bio_pair_release(bp); | |
504 | return 0; | |
505 | } | |
1da177e4 | 506 | |
fbb704ef | 507 | sector_offset = bio->bi_sector; |
508 | zone = find_zone(mddev->private, §or_offset); | |
509 | tmp_dev = map_sector(mddev, zone, bio->bi_sector, | |
510 | §or_offset); | |
1da177e4 | 511 | bio->bi_bdev = tmp_dev->bdev; |
fbb704ef | 512 | bio->bi_sector = sector_offset + zone->dev_start + |
513 | tmp_dev->data_offset; | |
1da177e4 LT |
514 | /* |
515 | * Let the main block layer submit the IO and resolve recursion: | |
516 | */ | |
517 | return 1; | |
518 | ||
519 | bad_map: | |
b5a20961 N |
520 | printk("md/raid0:%s: make_request bug: can't convert block across chunks" |
521 | " or bigger than %dk %llu %d\n", | |
522 | mdname(mddev), chunk_sects / 2, | |
523 | (unsigned long long)bio->bi_sector, bio->bi_size >> 10); | |
1da177e4 | 524 | |
6712ecf8 | 525 | bio_io_error(bio); |
1da177e4 LT |
526 | return 0; |
527 | } | |
8299d7f7 | 528 | |
1b961429 | 529 | static void raid0_status(struct seq_file *seq, mddev_t *mddev) |
1da177e4 LT |
530 | { |
531 | #undef MD_DEBUG | |
532 | #ifdef MD_DEBUG | |
533 | int j, k, h; | |
534 | char b[BDEVNAME_SIZE]; | |
070ec55d | 535 | raid0_conf_t *conf = mddev->private; |
84707f38 | 536 | int raid_disks = conf->strip_zone[0].nb_dev; |
8299d7f7 | 537 | |
1b961429 | 538 | sector_t zone_size; |
539 | sector_t zone_start = 0; | |
1da177e4 | 540 | h = 0; |
1b961429 | 541 | |
1da177e4 LT |
542 | for (j = 0; j < conf->nr_strip_zones; j++) { |
543 | seq_printf(seq, " z%d", j); | |
1da177e4 LT |
544 | seq_printf(seq, "=["); |
545 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
8299d7f7 | 546 | seq_printf(seq, "%s/", bdevname( |
84707f38 | 547 | conf->devlist[j*raid_disks + k] |
1b961429 | 548 | ->bdev, b)); |
549 | ||
550 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
551 | seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n", | |
552 | (unsigned long long)zone_start>>1, | |
553 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
554 | (unsigned long long)zone_size>>1); | |
555 | zone_start = conf->strip_zone[j].zone_end; | |
1da177e4 LT |
556 | } |
557 | #endif | |
9d8f0363 | 558 | seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); |
1da177e4 LT |
559 | return; |
560 | } | |
561 | ||
049d6c1e | 562 | static void *raid0_takeover_raid45(mddev_t *mddev) |
9af204cf TM |
563 | { |
564 | mdk_rdev_t *rdev; | |
565 | raid0_conf_t *priv_conf; | |
566 | ||
567 | if (mddev->degraded != 1) { | |
b5a20961 N |
568 | printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n", |
569 | mdname(mddev), | |
9af204cf TM |
570 | mddev->degraded); |
571 | return ERR_PTR(-EINVAL); | |
572 | } | |
573 | ||
574 | list_for_each_entry(rdev, &mddev->disks, same_set) { | |
575 | /* check slot number for a disk */ | |
576 | if (rdev->raid_disk == mddev->raid_disks-1) { | |
b5a20961 N |
577 | printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n", |
578 | mdname(mddev)); | |
9af204cf TM |
579 | return ERR_PTR(-EINVAL); |
580 | } | |
581 | } | |
582 | ||
583 | /* Set new parameters */ | |
584 | mddev->new_level = 0; | |
001048a3 | 585 | mddev->new_layout = 0; |
9af204cf TM |
586 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
587 | mddev->raid_disks--; | |
588 | mddev->delta_disks = -1; | |
589 | /* make sure it will be not marked as dirty */ | |
590 | mddev->recovery_cp = MaxSector; | |
591 | ||
592 | create_strip_zones(mddev, &priv_conf); | |
593 | return priv_conf; | |
594 | } | |
595 | ||
596 | static void *raid0_takeover_raid10(mddev_t *mddev) | |
597 | { | |
598 | raid0_conf_t *priv_conf; | |
599 | ||
600 | /* Check layout: | |
601 | * - far_copies must be 1 | |
602 | * - near_copies must be 2 | |
603 | * - disks number must be even | |
604 | * - all mirrors must be already degraded | |
605 | */ | |
606 | if (mddev->layout != ((1 << 8) + 2)) { | |
b5a20961 N |
607 | printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takover layout: 0x%x\n", |
608 | mdname(mddev), | |
9af204cf TM |
609 | mddev->layout); |
610 | return ERR_PTR(-EINVAL); | |
611 | } | |
612 | if (mddev->raid_disks & 1) { | |
b5a20961 N |
613 | printk(KERN_ERR "md/raid0:%s: Raid0 cannot takover Raid10 with odd disk number.\n", |
614 | mdname(mddev)); | |
9af204cf TM |
615 | return ERR_PTR(-EINVAL); |
616 | } | |
617 | if (mddev->degraded != (mddev->raid_disks>>1)) { | |
b5a20961 N |
618 | printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n", |
619 | mdname(mddev)); | |
9af204cf TM |
620 | return ERR_PTR(-EINVAL); |
621 | } | |
622 | ||
623 | /* Set new parameters */ | |
624 | mddev->new_level = 0; | |
001048a3 | 625 | mddev->new_layout = 0; |
9af204cf TM |
626 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
627 | mddev->delta_disks = - mddev->raid_disks / 2; | |
628 | mddev->raid_disks += mddev->delta_disks; | |
629 | mddev->degraded = 0; | |
630 | /* make sure it will be not marked as dirty */ | |
631 | mddev->recovery_cp = MaxSector; | |
632 | ||
633 | create_strip_zones(mddev, &priv_conf); | |
9af204cf TM |
634 | return priv_conf; |
635 | } | |
636 | ||
fc3a08b8 KW |
637 | static void *raid0_takeover_raid1(mddev_t *mddev) |
638 | { | |
639 | raid0_conf_t *priv_conf; | |
640 | ||
641 | /* Check layout: | |
642 | * - (N - 1) mirror drives must be already faulty | |
643 | */ | |
644 | if ((mddev->raid_disks - 1) != mddev->degraded) { | |
645 | printk(KERN_ERR "md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n", | |
646 | mdname(mddev)); | |
647 | return ERR_PTR(-EINVAL); | |
648 | } | |
649 | ||
650 | /* Set new parameters */ | |
651 | mddev->new_level = 0; | |
652 | mddev->new_layout = 0; | |
653 | mddev->new_chunk_sectors = 128; /* by default set chunk size to 64k */ | |
654 | mddev->delta_disks = 1 - mddev->raid_disks; | |
f7bee809 | 655 | mddev->raid_disks = 1; |
fc3a08b8 KW |
656 | /* make sure it will be not marked as dirty */ |
657 | mddev->recovery_cp = MaxSector; | |
658 | ||
659 | create_strip_zones(mddev, &priv_conf); | |
660 | return priv_conf; | |
661 | } | |
662 | ||
9af204cf TM |
663 | static void *raid0_takeover(mddev_t *mddev) |
664 | { | |
665 | /* raid0 can take over: | |
049d6c1e | 666 | * raid4 - if all data disks are active. |
9af204cf TM |
667 | * raid5 - providing it is Raid4 layout and one disk is faulty |
668 | * raid10 - assuming we have all necessary active disks | |
fc3a08b8 | 669 | * raid1 - with (N -1) mirror drives faulty |
9af204cf | 670 | */ |
049d6c1e MT |
671 | if (mddev->level == 4) |
672 | return raid0_takeover_raid45(mddev); | |
673 | ||
9af204cf TM |
674 | if (mddev->level == 5) { |
675 | if (mddev->layout == ALGORITHM_PARITY_N) | |
049d6c1e | 676 | return raid0_takeover_raid45(mddev); |
9af204cf | 677 | |
b5a20961 N |
678 | printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n", |
679 | mdname(mddev), ALGORITHM_PARITY_N); | |
9af204cf TM |
680 | } |
681 | ||
682 | if (mddev->level == 10) | |
683 | return raid0_takeover_raid10(mddev); | |
684 | ||
fc3a08b8 KW |
685 | if (mddev->level == 1) |
686 | return raid0_takeover_raid1(mddev); | |
687 | ||
688 | printk(KERN_ERR "Takeover from raid%i to raid0 not supported\n", | |
689 | mddev->level); | |
690 | ||
9af204cf TM |
691 | return ERR_PTR(-EINVAL); |
692 | } | |
693 | ||
694 | static void raid0_quiesce(mddev_t *mddev, int state) | |
695 | { | |
696 | } | |
697 | ||
2604b703 | 698 | static struct mdk_personality raid0_personality= |
1da177e4 LT |
699 | { |
700 | .name = "raid0", | |
2604b703 | 701 | .level = 0, |
1da177e4 LT |
702 | .owner = THIS_MODULE, |
703 | .make_request = raid0_make_request, | |
704 | .run = raid0_run, | |
705 | .stop = raid0_stop, | |
706 | .status = raid0_status, | |
80c3a6ce | 707 | .size = raid0_size, |
9af204cf TM |
708 | .takeover = raid0_takeover, |
709 | .quiesce = raid0_quiesce, | |
1da177e4 LT |
710 | }; |
711 | ||
712 | static int __init raid0_init (void) | |
713 | { | |
2604b703 | 714 | return register_md_personality (&raid0_personality); |
1da177e4 LT |
715 | } |
716 | ||
717 | static void raid0_exit (void) | |
718 | { | |
2604b703 | 719 | unregister_md_personality (&raid0_personality); |
1da177e4 LT |
720 | } |
721 | ||
722 | module_init(raid0_init); | |
723 | module_exit(raid0_exit); | |
724 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 725 | MODULE_DESCRIPTION("RAID0 (striping) personality for MD"); |
1da177e4 | 726 | MODULE_ALIAS("md-personality-2"); /* RAID0 */ |
d9d166c2 | 727 | MODULE_ALIAS("md-raid0"); |
2604b703 | 728 | MODULE_ALIAS("md-level-0"); |