]>
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> |
43b2e5d8 | 23 | #include "md.h" |
ef740c37 | 24 | #include "raid0.h" |
9af204cf | 25 | #include "raid5.h" |
1da177e4 | 26 | |
165125e1 | 27 | static void raid0_unplug(struct request_queue *q) |
1da177e4 LT |
28 | { |
29 | mddev_t *mddev = q->queuedata; | |
070ec55d | 30 | raid0_conf_t *conf = mddev->private; |
b414579f | 31 | mdk_rdev_t **devlist = conf->devlist; |
84707f38 | 32 | int raid_disks = conf->strip_zone[0].nb_dev; |
1da177e4 LT |
33 | int i; |
34 | ||
84707f38 | 35 | for (i=0; i < raid_disks; i++) { |
165125e1 | 36 | struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev); |
1da177e4 | 37 | |
2ad8b1ef | 38 | blk_unplug(r_queue); |
1da177e4 LT |
39 | } |
40 | } | |
41 | ||
26be34dc N |
42 | static int raid0_congested(void *data, int bits) |
43 | { | |
44 | mddev_t *mddev = data; | |
070ec55d | 45 | raid0_conf_t *conf = mddev->private; |
b414579f | 46 | mdk_rdev_t **devlist = conf->devlist; |
84707f38 | 47 | int raid_disks = conf->strip_zone[0].nb_dev; |
26be34dc N |
48 | int i, ret = 0; |
49 | ||
3fa841d7 N |
50 | if (mddev_congested(mddev, bits)) |
51 | return 1; | |
52 | ||
84707f38 | 53 | for (i = 0; i < raid_disks && !ret ; i++) { |
165125e1 | 54 | struct request_queue *q = bdev_get_queue(devlist[i]->bdev); |
26be34dc N |
55 | |
56 | ret |= bdi_congested(&q->backing_dev_info, bits); | |
57 | } | |
58 | return ret; | |
59 | } | |
60 | ||
46994191 | 61 | /* |
62 | * inform the user of the raid configuration | |
63 | */ | |
64 | static void dump_zones(mddev_t *mddev) | |
65 | { | |
66 | int j, k, h; | |
67 | sector_t zone_size = 0; | |
68 | sector_t zone_start = 0; | |
69 | char b[BDEVNAME_SIZE]; | |
70 | raid0_conf_t *conf = mddev->private; | |
84707f38 | 71 | int raid_disks = conf->strip_zone[0].nb_dev; |
46994191 | 72 | printk(KERN_INFO "******* %s configuration *********\n", |
73 | mdname(mddev)); | |
74 | h = 0; | |
75 | for (j = 0; j < conf->nr_strip_zones; j++) { | |
76 | printk(KERN_INFO "zone%d=[", j); | |
77 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
78 | printk("%s/", | |
84707f38 | 79 | bdevname(conf->devlist[j*raid_disks |
46994191 | 80 | + k]->bdev, b)); |
81 | printk("]\n"); | |
82 | ||
83 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
84 | printk(KERN_INFO " zone offset=%llukb " | |
85 | "device offset=%llukb size=%llukb\n", | |
86 | (unsigned long long)zone_start>>1, | |
87 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
88 | (unsigned long long)zone_size>>1); | |
89 | zone_start = conf->strip_zone[j].zone_end; | |
90 | } | |
91 | printk(KERN_INFO "**********************************\n\n"); | |
92 | } | |
93 | ||
9af204cf | 94 | static int create_strip_zones(mddev_t *mddev, raid0_conf_t **private_conf) |
1da177e4 | 95 | { |
a9f326eb | 96 | int i, c, err; |
49f357a2 | 97 | sector_t curr_zone_end, sectors; |
b414579f | 98 | mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev; |
1da177e4 LT |
99 | struct strip_zone *zone; |
100 | int cnt; | |
101 | char b[BDEVNAME_SIZE]; | |
ed7b0038 AN |
102 | raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL); |
103 | ||
104 | if (!conf) | |
105 | return -ENOMEM; | |
159ec1fc | 106 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
0825b87a | 107 | printk(KERN_INFO "raid0: looking at %s\n", |
1da177e4 LT |
108 | bdevname(rdev1->bdev,b)); |
109 | c = 0; | |
13f2682b N |
110 | |
111 | /* round size to chunk_size */ | |
112 | sectors = rdev1->sectors; | |
113 | sector_div(sectors, mddev->chunk_sectors); | |
114 | rdev1->sectors = sectors * mddev->chunk_sectors; | |
115 | ||
159ec1fc | 116 | list_for_each_entry(rdev2, &mddev->disks, same_set) { |
0825b87a | 117 | printk(KERN_INFO "raid0: comparing %s(%llu)", |
1da177e4 | 118 | bdevname(rdev1->bdev,b), |
dd8ac336 | 119 | (unsigned long long)rdev1->sectors); |
0825b87a | 120 | printk(KERN_INFO " with %s(%llu)\n", |
1da177e4 | 121 | bdevname(rdev2->bdev,b), |
dd8ac336 | 122 | (unsigned long long)rdev2->sectors); |
1da177e4 | 123 | if (rdev2 == rdev1) { |
0825b87a | 124 | printk(KERN_INFO "raid0: END\n"); |
1da177e4 LT |
125 | break; |
126 | } | |
dd8ac336 | 127 | if (rdev2->sectors == rdev1->sectors) { |
1da177e4 LT |
128 | /* |
129 | * Not unique, don't count it as a new | |
130 | * group | |
131 | */ | |
0825b87a | 132 | printk(KERN_INFO "raid0: EQUAL\n"); |
1da177e4 LT |
133 | c = 1; |
134 | break; | |
135 | } | |
0825b87a | 136 | printk(KERN_INFO "raid0: NOT EQUAL\n"); |
1da177e4 LT |
137 | } |
138 | if (!c) { | |
0825b87a | 139 | printk(KERN_INFO "raid0: ==> UNIQUE\n"); |
1da177e4 | 140 | conf->nr_strip_zones++; |
0825b87a AN |
141 | printk(KERN_INFO "raid0: %d zones\n", |
142 | conf->nr_strip_zones); | |
1da177e4 LT |
143 | } |
144 | } | |
0825b87a | 145 | printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones); |
ed7b0038 | 146 | err = -ENOMEM; |
9ffae0cf | 147 | conf->strip_zone = kzalloc(sizeof(struct strip_zone)* |
1da177e4 LT |
148 | conf->nr_strip_zones, GFP_KERNEL); |
149 | if (!conf->strip_zone) | |
ed7b0038 | 150 | goto abort; |
9ffae0cf | 151 | conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* |
1da177e4 LT |
152 | conf->nr_strip_zones*mddev->raid_disks, |
153 | GFP_KERNEL); | |
154 | if (!conf->devlist) | |
ed7b0038 | 155 | goto abort; |
1da177e4 | 156 | |
1da177e4 LT |
157 | /* The first zone must contain all devices, so here we check that |
158 | * there is a proper alignment of slots to devices and find them all | |
159 | */ | |
160 | zone = &conf->strip_zone[0]; | |
161 | cnt = 0; | |
162 | smallest = NULL; | |
b414579f | 163 | dev = conf->devlist; |
ed7b0038 | 164 | err = -EINVAL; |
159ec1fc | 165 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
1da177e4 LT |
166 | int j = rdev1->raid_disk; |
167 | ||
9af204cf TM |
168 | if (mddev->level == 10) |
169 | /* taking over a raid10-n2 array */ | |
170 | j /= 2; | |
171 | ||
1da177e4 | 172 | if (j < 0 || j >= mddev->raid_disks) { |
0825b87a AN |
173 | printk(KERN_ERR "raid0: bad disk number %d - " |
174 | "aborting!\n", j); | |
1da177e4 LT |
175 | goto abort; |
176 | } | |
b414579f | 177 | if (dev[j]) { |
0825b87a AN |
178 | printk(KERN_ERR "raid0: multiple devices for %d - " |
179 | "aborting!\n", j); | |
1da177e4 LT |
180 | goto abort; |
181 | } | |
b414579f | 182 | dev[j] = rdev1; |
1da177e4 | 183 | |
8f6c2e4b MP |
184 | disk_stack_limits(mddev->gendisk, rdev1->bdev, |
185 | rdev1->data_offset << 9); | |
1da177e4 | 186 | /* as we don't honour merge_bvec_fn, we must never risk |
627a2d3c N |
187 | * violating it, so limit ->max_segments to 1, lying within |
188 | * a single page. | |
1da177e4 LT |
189 | */ |
190 | ||
627a2d3c N |
191 | if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) { |
192 | blk_queue_max_segments(mddev->queue, 1); | |
193 | blk_queue_segment_boundary(mddev->queue, | |
194 | PAGE_CACHE_SIZE - 1); | |
195 | } | |
dd8ac336 | 196 | if (!smallest || (rdev1->sectors < smallest->sectors)) |
1da177e4 LT |
197 | smallest = rdev1; |
198 | cnt++; | |
199 | } | |
200 | if (cnt != mddev->raid_disks) { | |
0825b87a AN |
201 | printk(KERN_ERR "raid0: too few disks (%d of %d) - " |
202 | "aborting!\n", cnt, mddev->raid_disks); | |
1da177e4 LT |
203 | goto abort; |
204 | } | |
205 | zone->nb_dev = cnt; | |
49f357a2 | 206 | zone->zone_end = smallest->sectors * cnt; |
1da177e4 | 207 | |
49f357a2 | 208 | curr_zone_end = zone->zone_end; |
1da177e4 LT |
209 | |
210 | /* now do the other zones */ | |
211 | for (i = 1; i < conf->nr_strip_zones; i++) | |
212 | { | |
a9f326eb N |
213 | int j; |
214 | ||
1da177e4 | 215 | zone = conf->strip_zone + i; |
b414579f | 216 | dev = conf->devlist + i * mddev->raid_disks; |
1da177e4 | 217 | |
0825b87a | 218 | printk(KERN_INFO "raid0: zone %d\n", i); |
d27a43ab | 219 | zone->dev_start = smallest->sectors; |
1da177e4 LT |
220 | smallest = NULL; |
221 | c = 0; | |
222 | ||
223 | for (j=0; j<cnt; j++) { | |
b414579f | 224 | rdev = conf->devlist[j]; |
0825b87a AN |
225 | printk(KERN_INFO "raid0: checking %s ...", |
226 | bdevname(rdev->bdev, b)); | |
d27a43ab | 227 | if (rdev->sectors <= zone->dev_start) { |
0825b87a | 228 | printk(KERN_INFO " nope.\n"); |
dd8ac336 AN |
229 | continue; |
230 | } | |
231 | printk(KERN_INFO " contained as device %d\n", c); | |
b414579f | 232 | dev[c] = rdev; |
dd8ac336 AN |
233 | c++; |
234 | if (!smallest || rdev->sectors < smallest->sectors) { | |
235 | smallest = rdev; | |
236 | printk(KERN_INFO " (%llu) is smallest!.\n", | |
237 | (unsigned long long)rdev->sectors); | |
238 | } | |
1da177e4 LT |
239 | } |
240 | ||
241 | zone->nb_dev = c; | |
49f357a2 | 242 | sectors = (smallest->sectors - zone->dev_start) * c; |
83838ed8 | 243 | printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n", |
49f357a2 | 244 | zone->nb_dev, (unsigned long long)sectors); |
1da177e4 | 245 | |
49f357a2 | 246 | curr_zone_end += sectors; |
d27a43ab | 247 | zone->zone_end = curr_zone_end; |
1da177e4 | 248 | |
6b8796cc | 249 | printk(KERN_INFO "raid0: current zone start: %llu\n", |
d27a43ab | 250 | (unsigned long long)smallest->sectors); |
1da177e4 | 251 | } |
1da177e4 | 252 | mddev->queue->unplug_fn = raid0_unplug; |
26be34dc N |
253 | mddev->queue->backing_dev_info.congested_fn = raid0_congested; |
254 | mddev->queue->backing_dev_info.congested_data = mddev; | |
1da177e4 | 255 | |
92e59b6b | 256 | /* |
257 | * now since we have the hard sector sizes, we can make sure | |
258 | * chunk size is a multiple of that sector size | |
259 | */ | |
9d8f0363 | 260 | if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) { |
92e59b6b | 261 | printk(KERN_ERR "%s chunk_size of %d not valid\n", |
262 | mdname(mddev), | |
9d8f0363 | 263 | mddev->chunk_sectors << 9); |
92e59b6b | 264 | goto abort; |
265 | } | |
8f6c2e4b MP |
266 | |
267 | blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); | |
268 | blk_queue_io_opt(mddev->queue, | |
269 | (mddev->chunk_sectors << 9) * mddev->raid_disks); | |
270 | ||
0825b87a | 271 | printk(KERN_INFO "raid0: done.\n"); |
9af204cf TM |
272 | *private_conf = conf; |
273 | ||
1da177e4 | 274 | return 0; |
5568a603 | 275 | abort: |
ed7b0038 AN |
276 | kfree(conf->strip_zone); |
277 | kfree(conf->devlist); | |
278 | kfree(conf); | |
9af204cf | 279 | *private_conf = NULL; |
ed7b0038 | 280 | return err; |
1da177e4 LT |
281 | } |
282 | ||
283 | /** | |
284 | * raid0_mergeable_bvec -- tell bio layer if a two requests can be merged | |
285 | * @q: request queue | |
cc371e66 | 286 | * @bvm: properties of new bio |
1da177e4 LT |
287 | * @biovec: the request that could be merged to it. |
288 | * | |
289 | * Return amount of bytes we can accept at this offset | |
290 | */ | |
cc371e66 AK |
291 | static int raid0_mergeable_bvec(struct request_queue *q, |
292 | struct bvec_merge_data *bvm, | |
293 | struct bio_vec *biovec) | |
1da177e4 LT |
294 | { |
295 | mddev_t *mddev = q->queuedata; | |
cc371e66 | 296 | sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); |
1da177e4 | 297 | int max; |
9d8f0363 | 298 | unsigned int chunk_sectors = mddev->chunk_sectors; |
cc371e66 | 299 | unsigned int bio_sectors = bvm->bi_size >> 9; |
1da177e4 | 300 | |
d6e412ea | 301 | if (is_power_of_2(chunk_sectors)) |
fbb704ef | 302 | max = (chunk_sectors - ((sector & (chunk_sectors-1)) |
303 | + bio_sectors)) << 9; | |
304 | else | |
305 | max = (chunk_sectors - (sector_div(sector, chunk_sectors) | |
306 | + bio_sectors)) << 9; | |
1da177e4 LT |
307 | if (max < 0) max = 0; /* bio_add cannot handle a negative return */ |
308 | if (max <= biovec->bv_len && bio_sectors == 0) | |
309 | return biovec->bv_len; | |
310 | else | |
311 | return max; | |
312 | } | |
313 | ||
80c3a6ce DW |
314 | static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks) |
315 | { | |
316 | sector_t array_sectors = 0; | |
317 | mdk_rdev_t *rdev; | |
318 | ||
319 | WARN_ONCE(sectors || raid_disks, | |
320 | "%s does not support generic reshape\n", __func__); | |
321 | ||
322 | list_for_each_entry(rdev, &mddev->disks, same_set) | |
323 | array_sectors += rdev->sectors; | |
324 | ||
325 | return array_sectors; | |
326 | } | |
327 | ||
8f79cfcd | 328 | static int raid0_run(mddev_t *mddev) |
1da177e4 | 329 | { |
9af204cf | 330 | raid0_conf_t *conf; |
5568a603 | 331 | int ret; |
1da177e4 | 332 | |
9d8f0363 | 333 | if (mddev->chunk_sectors == 0) { |
fbb704ef | 334 | printk(KERN_ERR "md/raid0: chunk size must be set.\n"); |
2604b703 N |
335 | return -EINVAL; |
336 | } | |
0894cc30 AN |
337 | if (md_check_no_bitmap(mddev)) |
338 | return -EINVAL; | |
086fa5ff | 339 | blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); |
e7e72bf6 | 340 | mddev->queue->queue_lock = &mddev->queue->__queue_lock; |
1da177e4 | 341 | |
9af204cf TM |
342 | /* if private is not null, we are here after takeover */ |
343 | if (mddev->private == NULL) { | |
344 | ret = create_strip_zones(mddev, &conf); | |
345 | if (ret < 0) | |
346 | return ret; | |
347 | mddev->private = conf; | |
348 | } | |
349 | conf = mddev->private; | |
350 | if (conf->scale_raid_disks) { | |
351 | int i; | |
352 | for (i=0; i < conf->strip_zone[0].nb_dev; i++) | |
353 | conf->devlist[i]->raid_disk /= conf->scale_raid_disks; | |
354 | /* FIXME update sysfs rd links */ | |
355 | } | |
1da177e4 LT |
356 | |
357 | /* calculate array device size */ | |
1f403624 | 358 | md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); |
1da177e4 | 359 | |
ccacc7d2 AN |
360 | printk(KERN_INFO "raid0 : md_size is %llu sectors.\n", |
361 | (unsigned long long)mddev->array_sectors); | |
1da177e4 LT |
362 | /* calculate the max read-ahead size. |
363 | * For read-ahead of large files to be effective, we need to | |
364 | * readahead at least twice a whole stripe. i.e. number of devices | |
365 | * multiplied by chunk size times 2. | |
366 | * If an individual device has an ra_pages greater than the | |
367 | * chunk size, then we will not drive that device as hard as it | |
368 | * wants. We consider this a configuration error: a larger | |
369 | * chunksize should be used in that case. | |
370 | */ | |
371 | { | |
9d8f0363 AN |
372 | int stripe = mddev->raid_disks * |
373 | (mddev->chunk_sectors << 9) / PAGE_SIZE; | |
1da177e4 LT |
374 | if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) |
375 | mddev->queue->backing_dev_info.ra_pages = 2* stripe; | |
376 | } | |
377 | ||
1da177e4 | 378 | blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); |
46994191 | 379 | dump_zones(mddev); |
ac5e7113 | 380 | md_integrity_register(mddev); |
1da177e4 | 381 | return 0; |
1da177e4 LT |
382 | } |
383 | ||
fb5ab4b5 | 384 | static int raid0_stop(mddev_t *mddev) |
1da177e4 | 385 | { |
070ec55d | 386 | raid0_conf_t *conf = mddev->private; |
1da177e4 LT |
387 | |
388 | blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ | |
990a8baf | 389 | kfree(conf->strip_zone); |
fb5ab4b5 | 390 | kfree(conf->devlist); |
990a8baf | 391 | kfree(conf); |
1da177e4 | 392 | mddev->private = NULL; |
1da177e4 LT |
393 | return 0; |
394 | } | |
395 | ||
49f357a2 N |
396 | /* Find the zone which holds a particular offset |
397 | * Update *sectorp to be an offset in that zone | |
398 | */ | |
dc582663 | 399 | static struct strip_zone *find_zone(struct raid0_private_data *conf, |
49f357a2 | 400 | sector_t *sectorp) |
dc582663 AN |
401 | { |
402 | int i; | |
403 | struct strip_zone *z = conf->strip_zone; | |
49f357a2 | 404 | sector_t sector = *sectorp; |
dc582663 AN |
405 | |
406 | for (i = 0; i < conf->nr_strip_zones; i++) | |
49f357a2 N |
407 | if (sector < z[i].zone_end) { |
408 | if (i) | |
409 | *sectorp = sector - z[i-1].zone_end; | |
dc582663 | 410 | return z + i; |
49f357a2 | 411 | } |
dc582663 AN |
412 | BUG(); |
413 | } | |
414 | ||
fbb704ef | 415 | /* |
416 | * remaps the bio to the target device. we separate two flows. | |
417 | * power 2 flow and a general flow for the sake of perfromance | |
418 | */ | |
419 | static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone, | |
420 | sector_t sector, sector_t *sector_offset) | |
1da177e4 | 421 | { |
fbb704ef | 422 | unsigned int sect_in_chunk; |
423 | sector_t chunk; | |
070ec55d | 424 | raid0_conf_t *conf = mddev->private; |
84707f38 | 425 | int raid_disks = conf->strip_zone[0].nb_dev; |
9d8f0363 | 426 | unsigned int chunk_sects = mddev->chunk_sectors; |
fbb704ef | 427 | |
d6e412ea | 428 | if (is_power_of_2(chunk_sects)) { |
fbb704ef | 429 | int chunksect_bits = ffz(~chunk_sects); |
430 | /* find the sector offset inside the chunk */ | |
431 | sect_in_chunk = sector & (chunk_sects - 1); | |
432 | sector >>= chunksect_bits; | |
433 | /* chunk in zone */ | |
434 | chunk = *sector_offset; | |
435 | /* quotient is the chunk in real device*/ | |
436 | sector_div(chunk, zone->nb_dev << chunksect_bits); | |
437 | } else{ | |
438 | sect_in_chunk = sector_div(sector, chunk_sects); | |
439 | chunk = *sector_offset; | |
440 | sector_div(chunk, chunk_sects * zone->nb_dev); | |
441 | } | |
442 | /* | |
443 | * position the bio over the real device | |
444 | * real sector = chunk in device + starting of zone | |
445 | * + the position in the chunk | |
446 | */ | |
447 | *sector_offset = (chunk * chunk_sects) + sect_in_chunk; | |
84707f38 | 448 | return conf->devlist[(zone - conf->strip_zone)*raid_disks |
fbb704ef | 449 | + sector_div(sector, zone->nb_dev)]; |
450 | } | |
451 | ||
452 | /* | |
453 | * Is io distribute over 1 or more chunks ? | |
454 | */ | |
455 | static inline int is_io_in_chunk_boundary(mddev_t *mddev, | |
456 | unsigned int chunk_sects, struct bio *bio) | |
457 | { | |
d6e412ea | 458 | if (likely(is_power_of_2(chunk_sects))) { |
fbb704ef | 459 | return chunk_sects >= ((bio->bi_sector & (chunk_sects-1)) |
460 | + (bio->bi_size >> 9)); | |
461 | } else{ | |
462 | sector_t sector = bio->bi_sector; | |
463 | return chunk_sects >= (sector_div(sector, chunk_sects) | |
464 | + (bio->bi_size >> 9)); | |
465 | } | |
466 | } | |
467 | ||
468 | static int raid0_make_request(struct request_queue *q, struct bio *bio) | |
469 | { | |
470 | mddev_t *mddev = q->queuedata; | |
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 | |
1f98a13f | 476 | if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) { |
a2826aa9 | 477 | md_barrier_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)); | |
1da177e4 LT |
498 | if (raid0_make_request(q, &bp->bio1)) |
499 | generic_make_request(&bp->bio1); | |
500 | if (raid0_make_request(q, &bp->bio2)) | |
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: | |
520 | printk("raid0_make_request bug: can't convert block across chunks" | |
a4712005 | 521 | " or bigger than %dk %llu %d\n", chunk_sects / 2, |
1da177e4 LT |
522 | (unsigned long long)bio->bi_sector, bio->bi_size >> 10); |
523 | ||
6712ecf8 | 524 | bio_io_error(bio); |
1da177e4 LT |
525 | return 0; |
526 | } | |
8299d7f7 | 527 | |
1b961429 | 528 | static void raid0_status(struct seq_file *seq, mddev_t *mddev) |
1da177e4 LT |
529 | { |
530 | #undef MD_DEBUG | |
531 | #ifdef MD_DEBUG | |
532 | int j, k, h; | |
533 | char b[BDEVNAME_SIZE]; | |
070ec55d | 534 | raid0_conf_t *conf = mddev->private; |
84707f38 | 535 | int raid_disks = conf->strip_zone[0].nb_dev; |
8299d7f7 | 536 | |
1b961429 | 537 | sector_t zone_size; |
538 | sector_t zone_start = 0; | |
1da177e4 | 539 | h = 0; |
1b961429 | 540 | |
1da177e4 LT |
541 | for (j = 0; j < conf->nr_strip_zones; j++) { |
542 | seq_printf(seq, " z%d", j); | |
1da177e4 LT |
543 | seq_printf(seq, "=["); |
544 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
8299d7f7 | 545 | seq_printf(seq, "%s/", bdevname( |
84707f38 | 546 | conf->devlist[j*raid_disks + k] |
1b961429 | 547 | ->bdev, b)); |
548 | ||
549 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
550 | seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n", | |
551 | (unsigned long long)zone_start>>1, | |
552 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
553 | (unsigned long long)zone_size>>1); | |
554 | zone_start = conf->strip_zone[j].zone_end; | |
1da177e4 LT |
555 | } |
556 | #endif | |
9d8f0363 | 557 | seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); |
1da177e4 LT |
558 | return; |
559 | } | |
560 | ||
9af204cf TM |
561 | static void *raid0_takeover_raid5(mddev_t *mddev) |
562 | { | |
563 | mdk_rdev_t *rdev; | |
564 | raid0_conf_t *priv_conf; | |
565 | ||
566 | if (mddev->degraded != 1) { | |
567 | printk(KERN_ERR "md: raid5 must be degraded! Degraded disks: %d\n", | |
568 | mddev->degraded); | |
569 | return ERR_PTR(-EINVAL); | |
570 | } | |
571 | ||
572 | list_for_each_entry(rdev, &mddev->disks, same_set) { | |
573 | /* check slot number for a disk */ | |
574 | if (rdev->raid_disk == mddev->raid_disks-1) { | |
575 | printk(KERN_ERR "md: raid5 must have missing parity disk!\n"); | |
576 | return ERR_PTR(-EINVAL); | |
577 | } | |
578 | } | |
579 | ||
580 | /* Set new parameters */ | |
581 | mddev->new_level = 0; | |
582 | mddev->new_chunk_sectors = mddev->chunk_sectors; | |
583 | mddev->raid_disks--; | |
584 | mddev->delta_disks = -1; | |
585 | /* make sure it will be not marked as dirty */ | |
586 | mddev->recovery_cp = MaxSector; | |
587 | ||
588 | create_strip_zones(mddev, &priv_conf); | |
589 | return priv_conf; | |
590 | } | |
591 | ||
592 | static void *raid0_takeover_raid10(mddev_t *mddev) | |
593 | { | |
594 | raid0_conf_t *priv_conf; | |
595 | ||
596 | /* Check layout: | |
597 | * - far_copies must be 1 | |
598 | * - near_copies must be 2 | |
599 | * - disks number must be even | |
600 | * - all mirrors must be already degraded | |
601 | */ | |
602 | if (mddev->layout != ((1 << 8) + 2)) { | |
603 | printk(KERN_ERR "md: Raid0 cannot takover layout: %x\n", | |
604 | mddev->layout); | |
605 | return ERR_PTR(-EINVAL); | |
606 | } | |
607 | if (mddev->raid_disks & 1) { | |
608 | printk(KERN_ERR "md: Raid0 cannot takover Raid10 with odd disk number.\n"); | |
609 | return ERR_PTR(-EINVAL); | |
610 | } | |
611 | if (mddev->degraded != (mddev->raid_disks>>1)) { | |
612 | printk(KERN_ERR "md: All mirrors must be already degraded!\n"); | |
613 | return ERR_PTR(-EINVAL); | |
614 | } | |
615 | ||
616 | /* Set new parameters */ | |
617 | mddev->new_level = 0; | |
618 | mddev->new_chunk_sectors = mddev->chunk_sectors; | |
619 | mddev->delta_disks = - mddev->raid_disks / 2; | |
620 | mddev->raid_disks += mddev->delta_disks; | |
621 | mddev->degraded = 0; | |
622 | /* make sure it will be not marked as dirty */ | |
623 | mddev->recovery_cp = MaxSector; | |
624 | ||
625 | create_strip_zones(mddev, &priv_conf); | |
626 | priv_conf->scale_raid_disks = 2; | |
627 | return priv_conf; | |
628 | } | |
629 | ||
630 | static void *raid0_takeover(mddev_t *mddev) | |
631 | { | |
632 | /* raid0 can take over: | |
633 | * raid5 - providing it is Raid4 layout and one disk is faulty | |
634 | * raid10 - assuming we have all necessary active disks | |
635 | */ | |
636 | if (mddev->level == 5) { | |
637 | if (mddev->layout == ALGORITHM_PARITY_N) | |
638 | return raid0_takeover_raid5(mddev); | |
639 | ||
640 | printk(KERN_ERR "md: Raid can only takeover Raid5 with layout: %d\n", | |
641 | ALGORITHM_PARITY_N); | |
642 | } | |
643 | ||
644 | if (mddev->level == 10) | |
645 | return raid0_takeover_raid10(mddev); | |
646 | ||
647 | return ERR_PTR(-EINVAL); | |
648 | } | |
649 | ||
650 | static void raid0_quiesce(mddev_t *mddev, int state) | |
651 | { | |
652 | } | |
653 | ||
2604b703 | 654 | static struct mdk_personality raid0_personality= |
1da177e4 LT |
655 | { |
656 | .name = "raid0", | |
2604b703 | 657 | .level = 0, |
1da177e4 LT |
658 | .owner = THIS_MODULE, |
659 | .make_request = raid0_make_request, | |
660 | .run = raid0_run, | |
661 | .stop = raid0_stop, | |
662 | .status = raid0_status, | |
80c3a6ce | 663 | .size = raid0_size, |
9af204cf TM |
664 | .takeover = raid0_takeover, |
665 | .quiesce = raid0_quiesce, | |
1da177e4 LT |
666 | }; |
667 | ||
668 | static int __init raid0_init (void) | |
669 | { | |
2604b703 | 670 | return register_md_personality (&raid0_personality); |
1da177e4 LT |
671 | } |
672 | ||
673 | static void raid0_exit (void) | |
674 | { | |
2604b703 | 675 | unregister_md_personality (&raid0_personality); |
1da177e4 LT |
676 | } |
677 | ||
678 | module_init(raid0_init); | |
679 | module_exit(raid0_exit); | |
680 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 681 | MODULE_DESCRIPTION("RAID0 (striping) personality for MD"); |
1da177e4 | 682 | MODULE_ALIAS("md-personality-2"); /* RAID0 */ |
d9d166c2 | 683 | MODULE_ALIAS("md-raid0"); |
2604b703 | 684 | MODULE_ALIAS("md-level-0"); |