]>
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 | { | |
50de8df4 | 52 | int j, k; |
46994191 | 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; |
50de8df4 N |
58 | printk(KERN_INFO "md: RAID0 configuration for %s - %d zone%s\n", |
59 | mdname(mddev), | |
60 | conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s"); | |
46994191 | 61 | for (j = 0; j < conf->nr_strip_zones; j++) { |
50de8df4 | 62 | printk(KERN_INFO "md: zone%d=[", j); |
46994191 | 63 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) |
50de8df4 | 64 | printk(KERN_CONT "%s%s", k?"/":"", |
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; | |
50de8df4 N |
70 | printk(KERN_INFO " zone-offset=%10lluKB, " |
71 | "device-offset=%10lluKB, size=%10lluKB\n", | |
46994191 | 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 | } | |
50de8df4 | 77 | printk(KERN_INFO "\n"); |
46994191 | 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]; | |
50de8df4 | 88 | char b2[BDEVNAME_SIZE]; |
ed7b0038 AN |
89 | raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL); |
90 | ||
91 | if (!conf) | |
92 | return -ENOMEM; | |
159ec1fc | 93 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
50de8df4 N |
94 | pr_debug("md/raid0:%s: looking at %s\n", |
95 | mdname(mddev), | |
96 | bdevname(rdev1->bdev, b)); | |
1da177e4 | 97 | c = 0; |
13f2682b N |
98 | |
99 | /* round size to chunk_size */ | |
100 | sectors = rdev1->sectors; | |
101 | sector_div(sectors, mddev->chunk_sectors); | |
102 | rdev1->sectors = sectors * mddev->chunk_sectors; | |
103 | ||
159ec1fc | 104 | list_for_each_entry(rdev2, &mddev->disks, same_set) { |
50de8df4 N |
105 | pr_debug("md/raid0:%s: comparing %s(%llu)" |
106 | " with %s(%llu)\n", | |
107 | mdname(mddev), | |
108 | bdevname(rdev1->bdev,b), | |
109 | (unsigned long long)rdev1->sectors, | |
110 | bdevname(rdev2->bdev,b2), | |
111 | (unsigned long long)rdev2->sectors); | |
1da177e4 | 112 | if (rdev2 == rdev1) { |
50de8df4 N |
113 | pr_debug("md/raid0:%s: END\n", |
114 | mdname(mddev)); | |
1da177e4 LT |
115 | break; |
116 | } | |
dd8ac336 | 117 | if (rdev2->sectors == rdev1->sectors) { |
1da177e4 LT |
118 | /* |
119 | * Not unique, don't count it as a new | |
120 | * group | |
121 | */ | |
50de8df4 N |
122 | pr_debug("md/raid0:%s: EQUAL\n", |
123 | mdname(mddev)); | |
1da177e4 LT |
124 | c = 1; |
125 | break; | |
126 | } | |
50de8df4 N |
127 | pr_debug("md/raid0:%s: NOT EQUAL\n", |
128 | mdname(mddev)); | |
1da177e4 LT |
129 | } |
130 | if (!c) { | |
50de8df4 N |
131 | pr_debug("md/raid0:%s: ==> UNIQUE\n", |
132 | mdname(mddev)); | |
1da177e4 | 133 | conf->nr_strip_zones++; |
50de8df4 N |
134 | pr_debug("md/raid0:%s: %d zones\n", |
135 | mdname(mddev), conf->nr_strip_zones); | |
1da177e4 LT |
136 | } |
137 | } | |
50de8df4 N |
138 | pr_debug("md/raid0:%s: FINAL %d zones\n", |
139 | mdname(mddev), conf->nr_strip_zones); | |
ed7b0038 | 140 | err = -ENOMEM; |
9ffae0cf | 141 | conf->strip_zone = kzalloc(sizeof(struct strip_zone)* |
1da177e4 LT |
142 | conf->nr_strip_zones, GFP_KERNEL); |
143 | if (!conf->strip_zone) | |
ed7b0038 | 144 | goto abort; |
9ffae0cf | 145 | conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* |
1da177e4 LT |
146 | conf->nr_strip_zones*mddev->raid_disks, |
147 | GFP_KERNEL); | |
148 | if (!conf->devlist) | |
ed7b0038 | 149 | goto abort; |
1da177e4 | 150 | |
1da177e4 LT |
151 | /* The first zone must contain all devices, so here we check that |
152 | * there is a proper alignment of slots to devices and find them all | |
153 | */ | |
154 | zone = &conf->strip_zone[0]; | |
155 | cnt = 0; | |
156 | smallest = NULL; | |
b414579f | 157 | dev = conf->devlist; |
ed7b0038 | 158 | err = -EINVAL; |
159ec1fc | 159 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
1da177e4 LT |
160 | int j = rdev1->raid_disk; |
161 | ||
e93f68a1 | 162 | if (mddev->level == 10) { |
9af204cf TM |
163 | /* taking over a raid10-n2 array */ |
164 | j /= 2; | |
e93f68a1 N |
165 | rdev1->new_raid_disk = j; |
166 | } | |
9af204cf | 167 | |
fc3a08b8 KW |
168 | if (mddev->level == 1) { |
169 | /* taiking over a raid1 array- | |
170 | * we have only one active disk | |
171 | */ | |
172 | j = 0; | |
173 | rdev1->new_raid_disk = j; | |
174 | } | |
175 | ||
1da177e4 | 176 | if (j < 0 || j >= mddev->raid_disks) { |
b5a20961 N |
177 | printk(KERN_ERR "md/raid0:%s: bad disk number %d - " |
178 | "aborting!\n", mdname(mddev), j); | |
1da177e4 LT |
179 | goto abort; |
180 | } | |
b414579f | 181 | if (dev[j]) { |
b5a20961 N |
182 | printk(KERN_ERR "md/raid0:%s: multiple devices for %d - " |
183 | "aborting!\n", mdname(mddev), j); | |
1da177e4 LT |
184 | goto abort; |
185 | } | |
b414579f | 186 | dev[j] = rdev1; |
1da177e4 | 187 | |
8f6c2e4b MP |
188 | disk_stack_limits(mddev->gendisk, rdev1->bdev, |
189 | rdev1->data_offset << 9); | |
1da177e4 | 190 | /* as we don't honour merge_bvec_fn, we must never risk |
627a2d3c N |
191 | * violating it, so limit ->max_segments to 1, lying within |
192 | * a single page. | |
1da177e4 LT |
193 | */ |
194 | ||
627a2d3c N |
195 | if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) { |
196 | blk_queue_max_segments(mddev->queue, 1); | |
197 | blk_queue_segment_boundary(mddev->queue, | |
198 | PAGE_CACHE_SIZE - 1); | |
199 | } | |
dd8ac336 | 200 | if (!smallest || (rdev1->sectors < smallest->sectors)) |
1da177e4 LT |
201 | smallest = rdev1; |
202 | cnt++; | |
203 | } | |
204 | if (cnt != mddev->raid_disks) { | |
b5a20961 N |
205 | printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - " |
206 | "aborting!\n", mdname(mddev), cnt, mddev->raid_disks); | |
1da177e4 LT |
207 | goto abort; |
208 | } | |
209 | zone->nb_dev = cnt; | |
49f357a2 | 210 | zone->zone_end = smallest->sectors * cnt; |
1da177e4 | 211 | |
49f357a2 | 212 | curr_zone_end = zone->zone_end; |
1da177e4 LT |
213 | |
214 | /* now do the other zones */ | |
215 | for (i = 1; i < conf->nr_strip_zones; i++) | |
216 | { | |
a9f326eb N |
217 | int j; |
218 | ||
1da177e4 | 219 | zone = conf->strip_zone + i; |
b414579f | 220 | dev = conf->devlist + i * mddev->raid_disks; |
1da177e4 | 221 | |
50de8df4 | 222 | pr_debug("md/raid0:%s: zone %d\n", 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]; |
d27a43ab | 229 | if (rdev->sectors <= zone->dev_start) { |
50de8df4 N |
230 | pr_debug("md/raid0:%s: checking %s ... nope\n", |
231 | mdname(mddev), | |
232 | bdevname(rdev->bdev, b)); | |
dd8ac336 AN |
233 | continue; |
234 | } | |
50de8df4 N |
235 | pr_debug("md/raid0:%s: checking %s ..." |
236 | " contained as device %d\n", | |
237 | mdname(mddev), | |
238 | bdevname(rdev->bdev, b), c); | |
b414579f | 239 | dev[c] = rdev; |
dd8ac336 AN |
240 | c++; |
241 | if (!smallest || rdev->sectors < smallest->sectors) { | |
242 | smallest = rdev; | |
50de8df4 N |
243 | pr_debug("md/raid0:%s: (%llu) is smallest!.\n", |
244 | mdname(mddev), | |
245 | (unsigned long long)rdev->sectors); | |
dd8ac336 | 246 | } |
1da177e4 LT |
247 | } |
248 | ||
249 | zone->nb_dev = c; | |
49f357a2 | 250 | sectors = (smallest->sectors - zone->dev_start) * c; |
50de8df4 N |
251 | pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n", |
252 | mdname(mddev), | |
253 | zone->nb_dev, (unsigned long long)sectors); | |
1da177e4 | 254 | |
49f357a2 | 255 | curr_zone_end += sectors; |
d27a43ab | 256 | zone->zone_end = curr_zone_end; |
1da177e4 | 257 | |
50de8df4 N |
258 | pr_debug("md/raid0:%s: current zone start: %llu\n", |
259 | mdname(mddev), | |
260 | (unsigned long long)smallest->sectors); | |
1da177e4 | 261 | } |
26be34dc N |
262 | mddev->queue->backing_dev_info.congested_fn = raid0_congested; |
263 | mddev->queue->backing_dev_info.congested_data = mddev; | |
1da177e4 | 264 | |
92e59b6b | 265 | /* |
266 | * now since we have the hard sector sizes, we can make sure | |
267 | * chunk size is a multiple of that sector size | |
268 | */ | |
9d8f0363 | 269 | if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) { |
b5a20961 | 270 | printk(KERN_ERR "md/raid0:%s: chunk_size of %d not valid\n", |
92e59b6b | 271 | mdname(mddev), |
9d8f0363 | 272 | mddev->chunk_sectors << 9); |
92e59b6b | 273 | goto abort; |
274 | } | |
8f6c2e4b MP |
275 | |
276 | blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); | |
277 | blk_queue_io_opt(mddev->queue, | |
278 | (mddev->chunk_sectors << 9) * mddev->raid_disks); | |
279 | ||
50de8df4 | 280 | pr_debug("md/raid0:%s: done.\n", mdname(mddev)); |
9af204cf TM |
281 | *private_conf = conf; |
282 | ||
1da177e4 | 283 | return 0; |
5568a603 | 284 | abort: |
ed7b0038 AN |
285 | kfree(conf->strip_zone); |
286 | kfree(conf->devlist); | |
287 | kfree(conf); | |
9af204cf | 288 | *private_conf = NULL; |
ed7b0038 | 289 | return err; |
1da177e4 LT |
290 | } |
291 | ||
292 | /** | |
293 | * raid0_mergeable_bvec -- tell bio layer if a two requests can be merged | |
294 | * @q: request queue | |
cc371e66 | 295 | * @bvm: properties of new bio |
1da177e4 LT |
296 | * @biovec: the request that could be merged to it. |
297 | * | |
298 | * Return amount of bytes we can accept at this offset | |
299 | */ | |
cc371e66 AK |
300 | static int raid0_mergeable_bvec(struct request_queue *q, |
301 | struct bvec_merge_data *bvm, | |
302 | struct bio_vec *biovec) | |
1da177e4 LT |
303 | { |
304 | mddev_t *mddev = q->queuedata; | |
cc371e66 | 305 | sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); |
1da177e4 | 306 | int max; |
9d8f0363 | 307 | unsigned int chunk_sectors = mddev->chunk_sectors; |
cc371e66 | 308 | unsigned int bio_sectors = bvm->bi_size >> 9; |
1da177e4 | 309 | |
d6e412ea | 310 | if (is_power_of_2(chunk_sectors)) |
fbb704ef | 311 | max = (chunk_sectors - ((sector & (chunk_sectors-1)) |
312 | + bio_sectors)) << 9; | |
313 | else | |
314 | max = (chunk_sectors - (sector_div(sector, chunk_sectors) | |
315 | + bio_sectors)) << 9; | |
1da177e4 LT |
316 | if (max < 0) max = 0; /* bio_add cannot handle a negative return */ |
317 | if (max <= biovec->bv_len && bio_sectors == 0) | |
318 | return biovec->bv_len; | |
319 | else | |
320 | return max; | |
321 | } | |
322 | ||
80c3a6ce DW |
323 | static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks) |
324 | { | |
325 | sector_t array_sectors = 0; | |
326 | mdk_rdev_t *rdev; | |
327 | ||
328 | WARN_ONCE(sectors || raid_disks, | |
329 | "%s does not support generic reshape\n", __func__); | |
330 | ||
331 | list_for_each_entry(rdev, &mddev->disks, same_set) | |
332 | array_sectors += rdev->sectors; | |
333 | ||
334 | return array_sectors; | |
335 | } | |
336 | ||
8f79cfcd | 337 | static int raid0_run(mddev_t *mddev) |
1da177e4 | 338 | { |
9af204cf | 339 | raid0_conf_t *conf; |
5568a603 | 340 | int ret; |
1da177e4 | 341 | |
9d8f0363 | 342 | if (mddev->chunk_sectors == 0) { |
b5a20961 N |
343 | printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n", |
344 | mdname(mddev)); | |
2604b703 N |
345 | return -EINVAL; |
346 | } | |
0894cc30 AN |
347 | if (md_check_no_bitmap(mddev)) |
348 | return -EINVAL; | |
086fa5ff | 349 | blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); |
1da177e4 | 350 | |
9af204cf TM |
351 | /* if private is not null, we are here after takeover */ |
352 | if (mddev->private == NULL) { | |
353 | ret = create_strip_zones(mddev, &conf); | |
354 | if (ret < 0) | |
355 | return ret; | |
356 | mddev->private = conf; | |
357 | } | |
358 | conf = mddev->private; | |
1da177e4 LT |
359 | |
360 | /* calculate array device size */ | |
1f403624 | 361 | md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); |
1da177e4 | 362 | |
b5a20961 N |
363 | printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n", |
364 | mdname(mddev), | |
365 | (unsigned long long)mddev->array_sectors); | |
1da177e4 LT |
366 | /* calculate the max read-ahead size. |
367 | * For read-ahead of large files to be effective, we need to | |
368 | * readahead at least twice a whole stripe. i.e. number of devices | |
369 | * multiplied by chunk size times 2. | |
370 | * If an individual device has an ra_pages greater than the | |
371 | * chunk size, then we will not drive that device as hard as it | |
372 | * wants. We consider this a configuration error: a larger | |
373 | * chunksize should be used in that case. | |
374 | */ | |
375 | { | |
9d8f0363 AN |
376 | int stripe = mddev->raid_disks * |
377 | (mddev->chunk_sectors << 9) / PAGE_SIZE; | |
1da177e4 LT |
378 | if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) |
379 | mddev->queue->backing_dev_info.ra_pages = 2* stripe; | |
380 | } | |
381 | ||
1da177e4 | 382 | blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); |
46994191 | 383 | dump_zones(mddev); |
a91a2785 | 384 | return md_integrity_register(mddev); |
1da177e4 LT |
385 | } |
386 | ||
fb5ab4b5 | 387 | static int raid0_stop(mddev_t *mddev) |
1da177e4 | 388 | { |
070ec55d | 389 | raid0_conf_t *conf = mddev->private; |
1da177e4 LT |
390 | |
391 | blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ | |
990a8baf | 392 | kfree(conf->strip_zone); |
fb5ab4b5 | 393 | kfree(conf->devlist); |
990a8baf | 394 | kfree(conf); |
1da177e4 | 395 | mddev->private = NULL; |
1da177e4 LT |
396 | return 0; |
397 | } | |
398 | ||
49f357a2 N |
399 | /* Find the zone which holds a particular offset |
400 | * Update *sectorp to be an offset in that zone | |
401 | */ | |
dc582663 | 402 | static struct strip_zone *find_zone(struct raid0_private_data *conf, |
49f357a2 | 403 | sector_t *sectorp) |
dc582663 AN |
404 | { |
405 | int i; | |
406 | struct strip_zone *z = conf->strip_zone; | |
49f357a2 | 407 | sector_t sector = *sectorp; |
dc582663 AN |
408 | |
409 | for (i = 0; i < conf->nr_strip_zones; i++) | |
49f357a2 N |
410 | if (sector < z[i].zone_end) { |
411 | if (i) | |
412 | *sectorp = sector - z[i-1].zone_end; | |
dc582663 | 413 | return z + i; |
49f357a2 | 414 | } |
dc582663 AN |
415 | BUG(); |
416 | } | |
417 | ||
fbb704ef | 418 | /* |
419 | * remaps the bio to the target device. we separate two flows. | |
420 | * power 2 flow and a general flow for the sake of perfromance | |
421 | */ | |
422 | static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone, | |
423 | sector_t sector, sector_t *sector_offset) | |
1da177e4 | 424 | { |
fbb704ef | 425 | unsigned int sect_in_chunk; |
426 | sector_t chunk; | |
070ec55d | 427 | raid0_conf_t *conf = mddev->private; |
84707f38 | 428 | int raid_disks = conf->strip_zone[0].nb_dev; |
9d8f0363 | 429 | unsigned int chunk_sects = mddev->chunk_sectors; |
fbb704ef | 430 | |
d6e412ea | 431 | if (is_power_of_2(chunk_sects)) { |
fbb704ef | 432 | int chunksect_bits = ffz(~chunk_sects); |
433 | /* find the sector offset inside the chunk */ | |
434 | sect_in_chunk = sector & (chunk_sects - 1); | |
435 | sector >>= chunksect_bits; | |
436 | /* chunk in zone */ | |
437 | chunk = *sector_offset; | |
438 | /* quotient is the chunk in real device*/ | |
439 | sector_div(chunk, zone->nb_dev << chunksect_bits); | |
440 | } else{ | |
441 | sect_in_chunk = sector_div(sector, chunk_sects); | |
442 | chunk = *sector_offset; | |
443 | sector_div(chunk, chunk_sects * zone->nb_dev); | |
444 | } | |
445 | /* | |
446 | * position the bio over the real device | |
447 | * real sector = chunk in device + starting of zone | |
448 | * + the position in the chunk | |
449 | */ | |
450 | *sector_offset = (chunk * chunk_sects) + sect_in_chunk; | |
84707f38 | 451 | return conf->devlist[(zone - conf->strip_zone)*raid_disks |
fbb704ef | 452 | + sector_div(sector, zone->nb_dev)]; |
453 | } | |
454 | ||
455 | /* | |
456 | * Is io distribute over 1 or more chunks ? | |
457 | */ | |
458 | static inline int is_io_in_chunk_boundary(mddev_t *mddev, | |
459 | unsigned int chunk_sects, struct bio *bio) | |
460 | { | |
d6e412ea | 461 | if (likely(is_power_of_2(chunk_sects))) { |
fbb704ef | 462 | return chunk_sects >= ((bio->bi_sector & (chunk_sects-1)) |
463 | + (bio->bi_size >> 9)); | |
464 | } else{ | |
465 | sector_t sector = bio->bi_sector; | |
466 | return chunk_sects >= (sector_div(sector, chunk_sects) | |
467 | + (bio->bi_size >> 9)); | |
468 | } | |
469 | } | |
470 | ||
21a52c6d | 471 | static int raid0_make_request(mddev_t *mddev, struct bio *bio) |
fbb704ef | 472 | { |
fbb704ef | 473 | unsigned int chunk_sects; |
474 | sector_t sector_offset; | |
1da177e4 LT |
475 | struct strip_zone *zone; |
476 | mdk_rdev_t *tmp_dev; | |
1da177e4 | 477 | |
e9c7469b TH |
478 | if (unlikely(bio->bi_rw & REQ_FLUSH)) { |
479 | md_flush_request(mddev, bio); | |
e5dcdd80 N |
480 | return 0; |
481 | } | |
482 | ||
9d8f0363 | 483 | chunk_sects = mddev->chunk_sectors; |
fbb704ef | 484 | if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) { |
485 | sector_t sector = bio->bi_sector; | |
1da177e4 LT |
486 | struct bio_pair *bp; |
487 | /* Sanity check -- queue functions should prevent this happening */ | |
488 | if (bio->bi_vcnt != 1 || | |
489 | bio->bi_idx != 0) | |
490 | goto bad_map; | |
491 | /* This is a one page bio that upper layers | |
492 | * refuse to split for us, so we need to split it. | |
493 | */ | |
d6e412ea | 494 | if (likely(is_power_of_2(chunk_sects))) |
fbb704ef | 495 | bp = bio_split(bio, chunk_sects - (sector & |
496 | (chunk_sects-1))); | |
497 | else | |
498 | bp = bio_split(bio, chunk_sects - | |
499 | sector_div(sector, chunk_sects)); | |
21a52c6d | 500 | if (raid0_make_request(mddev, &bp->bio1)) |
1da177e4 | 501 | generic_make_request(&bp->bio1); |
21a52c6d | 502 | if (raid0_make_request(mddev, &bp->bio2)) |
1da177e4 LT |
503 | generic_make_request(&bp->bio2); |
504 | ||
505 | bio_pair_release(bp); | |
506 | return 0; | |
507 | } | |
1da177e4 | 508 | |
fbb704ef | 509 | sector_offset = bio->bi_sector; |
510 | zone = find_zone(mddev->private, §or_offset); | |
511 | tmp_dev = map_sector(mddev, zone, bio->bi_sector, | |
512 | §or_offset); | |
1da177e4 | 513 | bio->bi_bdev = tmp_dev->bdev; |
fbb704ef | 514 | bio->bi_sector = sector_offset + zone->dev_start + |
515 | tmp_dev->data_offset; | |
1da177e4 LT |
516 | /* |
517 | * Let the main block layer submit the IO and resolve recursion: | |
518 | */ | |
519 | return 1; | |
520 | ||
521 | bad_map: | |
b5a20961 N |
522 | printk("md/raid0:%s: make_request bug: can't convert block across chunks" |
523 | " or bigger than %dk %llu %d\n", | |
524 | mdname(mddev), chunk_sects / 2, | |
525 | (unsigned long long)bio->bi_sector, bio->bi_size >> 10); | |
1da177e4 | 526 | |
6712ecf8 | 527 | bio_io_error(bio); |
1da177e4 LT |
528 | return 0; |
529 | } | |
8299d7f7 | 530 | |
1b961429 | 531 | static void raid0_status(struct seq_file *seq, mddev_t *mddev) |
1da177e4 | 532 | { |
9d8f0363 | 533 | seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); |
1da177e4 LT |
534 | return; |
535 | } | |
536 | ||
049d6c1e | 537 | static void *raid0_takeover_raid45(mddev_t *mddev) |
9af204cf TM |
538 | { |
539 | mdk_rdev_t *rdev; | |
540 | raid0_conf_t *priv_conf; | |
541 | ||
542 | if (mddev->degraded != 1) { | |
b5a20961 N |
543 | printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n", |
544 | mdname(mddev), | |
9af204cf TM |
545 | mddev->degraded); |
546 | return ERR_PTR(-EINVAL); | |
547 | } | |
548 | ||
549 | list_for_each_entry(rdev, &mddev->disks, same_set) { | |
550 | /* check slot number for a disk */ | |
551 | if (rdev->raid_disk == mddev->raid_disks-1) { | |
b5a20961 N |
552 | printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n", |
553 | mdname(mddev)); | |
9af204cf TM |
554 | return ERR_PTR(-EINVAL); |
555 | } | |
556 | } | |
557 | ||
558 | /* Set new parameters */ | |
559 | mddev->new_level = 0; | |
001048a3 | 560 | mddev->new_layout = 0; |
9af204cf TM |
561 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
562 | mddev->raid_disks--; | |
563 | mddev->delta_disks = -1; | |
564 | /* make sure it will be not marked as dirty */ | |
565 | mddev->recovery_cp = MaxSector; | |
566 | ||
567 | create_strip_zones(mddev, &priv_conf); | |
568 | return priv_conf; | |
569 | } | |
570 | ||
571 | static void *raid0_takeover_raid10(mddev_t *mddev) | |
572 | { | |
573 | raid0_conf_t *priv_conf; | |
574 | ||
575 | /* Check layout: | |
576 | * - far_copies must be 1 | |
577 | * - near_copies must be 2 | |
578 | * - disks number must be even | |
579 | * - all mirrors must be already degraded | |
580 | */ | |
581 | if (mddev->layout != ((1 << 8) + 2)) { | |
b5a20961 N |
582 | printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takover layout: 0x%x\n", |
583 | mdname(mddev), | |
9af204cf TM |
584 | mddev->layout); |
585 | return ERR_PTR(-EINVAL); | |
586 | } | |
587 | if (mddev->raid_disks & 1) { | |
b5a20961 N |
588 | printk(KERN_ERR "md/raid0:%s: Raid0 cannot takover Raid10 with odd disk number.\n", |
589 | mdname(mddev)); | |
9af204cf TM |
590 | return ERR_PTR(-EINVAL); |
591 | } | |
592 | if (mddev->degraded != (mddev->raid_disks>>1)) { | |
b5a20961 N |
593 | printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n", |
594 | mdname(mddev)); | |
9af204cf TM |
595 | return ERR_PTR(-EINVAL); |
596 | } | |
597 | ||
598 | /* Set new parameters */ | |
599 | mddev->new_level = 0; | |
001048a3 | 600 | mddev->new_layout = 0; |
9af204cf TM |
601 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
602 | mddev->delta_disks = - mddev->raid_disks / 2; | |
603 | mddev->raid_disks += mddev->delta_disks; | |
604 | mddev->degraded = 0; | |
605 | /* make sure it will be not marked as dirty */ | |
606 | mddev->recovery_cp = MaxSector; | |
607 | ||
608 | create_strip_zones(mddev, &priv_conf); | |
9af204cf TM |
609 | return priv_conf; |
610 | } | |
611 | ||
fc3a08b8 KW |
612 | static void *raid0_takeover_raid1(mddev_t *mddev) |
613 | { | |
614 | raid0_conf_t *priv_conf; | |
615 | ||
616 | /* Check layout: | |
617 | * - (N - 1) mirror drives must be already faulty | |
618 | */ | |
619 | if ((mddev->raid_disks - 1) != mddev->degraded) { | |
620 | printk(KERN_ERR "md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n", | |
621 | mdname(mddev)); | |
622 | return ERR_PTR(-EINVAL); | |
623 | } | |
624 | ||
625 | /* Set new parameters */ | |
626 | mddev->new_level = 0; | |
627 | mddev->new_layout = 0; | |
628 | mddev->new_chunk_sectors = 128; /* by default set chunk size to 64k */ | |
629 | mddev->delta_disks = 1 - mddev->raid_disks; | |
f7bee809 | 630 | mddev->raid_disks = 1; |
fc3a08b8 KW |
631 | /* make sure it will be not marked as dirty */ |
632 | mddev->recovery_cp = MaxSector; | |
633 | ||
634 | create_strip_zones(mddev, &priv_conf); | |
635 | return priv_conf; | |
636 | } | |
637 | ||
9af204cf TM |
638 | static void *raid0_takeover(mddev_t *mddev) |
639 | { | |
640 | /* raid0 can take over: | |
049d6c1e | 641 | * raid4 - if all data disks are active. |
9af204cf TM |
642 | * raid5 - providing it is Raid4 layout and one disk is faulty |
643 | * raid10 - assuming we have all necessary active disks | |
fc3a08b8 | 644 | * raid1 - with (N -1) mirror drives faulty |
9af204cf | 645 | */ |
049d6c1e MT |
646 | if (mddev->level == 4) |
647 | return raid0_takeover_raid45(mddev); | |
648 | ||
9af204cf TM |
649 | if (mddev->level == 5) { |
650 | if (mddev->layout == ALGORITHM_PARITY_N) | |
049d6c1e | 651 | return raid0_takeover_raid45(mddev); |
9af204cf | 652 | |
b5a20961 N |
653 | printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n", |
654 | mdname(mddev), ALGORITHM_PARITY_N); | |
9af204cf TM |
655 | } |
656 | ||
657 | if (mddev->level == 10) | |
658 | return raid0_takeover_raid10(mddev); | |
659 | ||
fc3a08b8 KW |
660 | if (mddev->level == 1) |
661 | return raid0_takeover_raid1(mddev); | |
662 | ||
663 | printk(KERN_ERR "Takeover from raid%i to raid0 not supported\n", | |
664 | mddev->level); | |
665 | ||
9af204cf TM |
666 | return ERR_PTR(-EINVAL); |
667 | } | |
668 | ||
669 | static void raid0_quiesce(mddev_t *mddev, int state) | |
670 | { | |
671 | } | |
672 | ||
2604b703 | 673 | static struct mdk_personality raid0_personality= |
1da177e4 LT |
674 | { |
675 | .name = "raid0", | |
2604b703 | 676 | .level = 0, |
1da177e4 LT |
677 | .owner = THIS_MODULE, |
678 | .make_request = raid0_make_request, | |
679 | .run = raid0_run, | |
680 | .stop = raid0_stop, | |
681 | .status = raid0_status, | |
80c3a6ce | 682 | .size = raid0_size, |
9af204cf TM |
683 | .takeover = raid0_takeover, |
684 | .quiesce = raid0_quiesce, | |
1da177e4 LT |
685 | }; |
686 | ||
687 | static int __init raid0_init (void) | |
688 | { | |
2604b703 | 689 | return register_md_personality (&raid0_personality); |
1da177e4 LT |
690 | } |
691 | ||
692 | static void raid0_exit (void) | |
693 | { | |
2604b703 | 694 | unregister_md_personality (&raid0_personality); |
1da177e4 LT |
695 | } |
696 | ||
697 | module_init(raid0_init); | |
698 | module_exit(raid0_exit); | |
699 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 700 | MODULE_DESCRIPTION("RAID0 (striping) personality for MD"); |
1da177e4 | 701 | MODULE_ALIAS("md-personality-2"); /* RAID0 */ |
d9d166c2 | 702 | MODULE_ALIAS("md-raid0"); |
2604b703 | 703 | MODULE_ALIAS("md-level-0"); |