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" |
1da177e4 | 25 | |
165125e1 | 26 | static void raid0_unplug(struct request_queue *q) |
1da177e4 LT |
27 | { |
28 | mddev_t *mddev = q->queuedata; | |
070ec55d | 29 | raid0_conf_t *conf = mddev->private; |
b414579f | 30 | mdk_rdev_t **devlist = conf->devlist; |
84707f38 | 31 | int raid_disks = conf->strip_zone[0].nb_dev; |
1da177e4 LT |
32 | int i; |
33 | ||
84707f38 | 34 | for (i=0; i < raid_disks; i++) { |
165125e1 | 35 | struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev); |
1da177e4 | 36 | |
2ad8b1ef | 37 | blk_unplug(r_queue); |
1da177e4 LT |
38 | } |
39 | } | |
40 | ||
26be34dc N |
41 | static int raid0_congested(void *data, int bits) |
42 | { | |
43 | mddev_t *mddev = data; | |
070ec55d | 44 | raid0_conf_t *conf = mddev->private; |
b414579f | 45 | mdk_rdev_t **devlist = conf->devlist; |
84707f38 | 46 | int raid_disks = conf->strip_zone[0].nb_dev; |
26be34dc N |
47 | int i, ret = 0; |
48 | ||
3fa841d7 N |
49 | if (mddev_congested(mddev, bits)) |
50 | return 1; | |
51 | ||
84707f38 | 52 | for (i = 0; i < raid_disks && !ret ; i++) { |
165125e1 | 53 | struct request_queue *q = bdev_get_queue(devlist[i]->bdev); |
26be34dc N |
54 | |
55 | ret |= bdi_congested(&q->backing_dev_info, bits); | |
56 | } | |
57 | return ret; | |
58 | } | |
59 | ||
46994191 | 60 | /* |
61 | * inform the user of the raid configuration | |
62 | */ | |
63 | static void dump_zones(mddev_t *mddev) | |
64 | { | |
65 | int j, k, h; | |
66 | sector_t zone_size = 0; | |
67 | sector_t zone_start = 0; | |
68 | char b[BDEVNAME_SIZE]; | |
69 | raid0_conf_t *conf = mddev->private; | |
84707f38 | 70 | int raid_disks = conf->strip_zone[0].nb_dev; |
46994191 | 71 | printk(KERN_INFO "******* %s configuration *********\n", |
72 | mdname(mddev)); | |
73 | h = 0; | |
74 | for (j = 0; j < conf->nr_strip_zones; j++) { | |
75 | printk(KERN_INFO "zone%d=[", j); | |
76 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
77 | printk("%s/", | |
84707f38 | 78 | bdevname(conf->devlist[j*raid_disks |
46994191 | 79 | + k]->bdev, b)); |
80 | printk("]\n"); | |
81 | ||
82 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
83 | printk(KERN_INFO " zone offset=%llukb " | |
84 | "device offset=%llukb size=%llukb\n", | |
85 | (unsigned long long)zone_start>>1, | |
86 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
87 | (unsigned long long)zone_size>>1); | |
88 | zone_start = conf->strip_zone[j].zone_end; | |
89 | } | |
90 | printk(KERN_INFO "**********************************\n\n"); | |
91 | } | |
92 | ||
ed7b0038 | 93 | static int create_strip_zones(mddev_t *mddev) |
1da177e4 | 94 | { |
a9f326eb | 95 | int i, c, err; |
49f357a2 | 96 | sector_t curr_zone_end, sectors; |
b414579f | 97 | mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev; |
1da177e4 LT |
98 | struct strip_zone *zone; |
99 | int cnt; | |
100 | char b[BDEVNAME_SIZE]; | |
ed7b0038 AN |
101 | raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL); |
102 | ||
103 | if (!conf) | |
104 | return -ENOMEM; | |
159ec1fc | 105 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
0825b87a | 106 | printk(KERN_INFO "raid0: looking at %s\n", |
1da177e4 LT |
107 | bdevname(rdev1->bdev,b)); |
108 | c = 0; | |
13f2682b N |
109 | |
110 | /* round size to chunk_size */ | |
111 | sectors = rdev1->sectors; | |
112 | sector_div(sectors, mddev->chunk_sectors); | |
113 | rdev1->sectors = sectors * mddev->chunk_sectors; | |
114 | ||
159ec1fc | 115 | list_for_each_entry(rdev2, &mddev->disks, same_set) { |
0825b87a | 116 | printk(KERN_INFO "raid0: comparing %s(%llu)", |
1da177e4 | 117 | bdevname(rdev1->bdev,b), |
dd8ac336 | 118 | (unsigned long long)rdev1->sectors); |
0825b87a | 119 | printk(KERN_INFO " with %s(%llu)\n", |
1da177e4 | 120 | bdevname(rdev2->bdev,b), |
dd8ac336 | 121 | (unsigned long long)rdev2->sectors); |
1da177e4 | 122 | if (rdev2 == rdev1) { |
0825b87a | 123 | printk(KERN_INFO "raid0: END\n"); |
1da177e4 LT |
124 | break; |
125 | } | |
dd8ac336 | 126 | if (rdev2->sectors == rdev1->sectors) { |
1da177e4 LT |
127 | /* |
128 | * Not unique, don't count it as a new | |
129 | * group | |
130 | */ | |
0825b87a | 131 | printk(KERN_INFO "raid0: EQUAL\n"); |
1da177e4 LT |
132 | c = 1; |
133 | break; | |
134 | } | |
0825b87a | 135 | printk(KERN_INFO "raid0: NOT EQUAL\n"); |
1da177e4 LT |
136 | } |
137 | if (!c) { | |
0825b87a | 138 | printk(KERN_INFO "raid0: ==> UNIQUE\n"); |
1da177e4 | 139 | conf->nr_strip_zones++; |
0825b87a AN |
140 | printk(KERN_INFO "raid0: %d zones\n", |
141 | conf->nr_strip_zones); | |
1da177e4 LT |
142 | } |
143 | } | |
0825b87a | 144 | printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones); |
ed7b0038 | 145 | err = -ENOMEM; |
9ffae0cf | 146 | conf->strip_zone = kzalloc(sizeof(struct strip_zone)* |
1da177e4 LT |
147 | conf->nr_strip_zones, GFP_KERNEL); |
148 | if (!conf->strip_zone) | |
ed7b0038 | 149 | goto abort; |
9ffae0cf | 150 | conf->devlist = kzalloc(sizeof(mdk_rdev_t*)* |
1da177e4 LT |
151 | conf->nr_strip_zones*mddev->raid_disks, |
152 | GFP_KERNEL); | |
153 | if (!conf->devlist) | |
ed7b0038 | 154 | goto abort; |
1da177e4 | 155 | |
1da177e4 LT |
156 | /* The first zone must contain all devices, so here we check that |
157 | * there is a proper alignment of slots to devices and find them all | |
158 | */ | |
159 | zone = &conf->strip_zone[0]; | |
160 | cnt = 0; | |
161 | smallest = NULL; | |
b414579f | 162 | dev = conf->devlist; |
ed7b0038 | 163 | err = -EINVAL; |
159ec1fc | 164 | list_for_each_entry(rdev1, &mddev->disks, same_set) { |
1da177e4 LT |
165 | int j = rdev1->raid_disk; |
166 | ||
167 | if (j < 0 || j >= mddev->raid_disks) { | |
0825b87a AN |
168 | printk(KERN_ERR "raid0: bad disk number %d - " |
169 | "aborting!\n", j); | |
1da177e4 LT |
170 | goto abort; |
171 | } | |
b414579f | 172 | if (dev[j]) { |
0825b87a AN |
173 | printk(KERN_ERR "raid0: multiple devices for %d - " |
174 | "aborting!\n", j); | |
1da177e4 LT |
175 | goto abort; |
176 | } | |
b414579f | 177 | dev[j] = rdev1; |
1da177e4 | 178 | |
8f6c2e4b MP |
179 | disk_stack_limits(mddev->gendisk, rdev1->bdev, |
180 | rdev1->data_offset << 9); | |
1da177e4 | 181 | /* as we don't honour merge_bvec_fn, we must never risk |
627a2d3c N |
182 | * violating it, so limit ->max_segments to 1, lying within |
183 | * a single page. | |
1da177e4 LT |
184 | */ |
185 | ||
627a2d3c N |
186 | if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) { |
187 | blk_queue_max_segments(mddev->queue, 1); | |
188 | blk_queue_segment_boundary(mddev->queue, | |
189 | PAGE_CACHE_SIZE - 1); | |
190 | } | |
dd8ac336 | 191 | if (!smallest || (rdev1->sectors < smallest->sectors)) |
1da177e4 LT |
192 | smallest = rdev1; |
193 | cnt++; | |
194 | } | |
195 | if (cnt != mddev->raid_disks) { | |
0825b87a AN |
196 | printk(KERN_ERR "raid0: too few disks (%d of %d) - " |
197 | "aborting!\n", cnt, mddev->raid_disks); | |
1da177e4 LT |
198 | goto abort; |
199 | } | |
200 | zone->nb_dev = cnt; | |
49f357a2 | 201 | zone->zone_end = smallest->sectors * cnt; |
1da177e4 | 202 | |
49f357a2 | 203 | curr_zone_end = zone->zone_end; |
1da177e4 LT |
204 | |
205 | /* now do the other zones */ | |
206 | for (i = 1; i < conf->nr_strip_zones; i++) | |
207 | { | |
a9f326eb N |
208 | int j; |
209 | ||
1da177e4 | 210 | zone = conf->strip_zone + i; |
b414579f | 211 | dev = conf->devlist + i * mddev->raid_disks; |
1da177e4 | 212 | |
0825b87a | 213 | printk(KERN_INFO "raid0: zone %d\n", i); |
d27a43ab | 214 | zone->dev_start = smallest->sectors; |
1da177e4 LT |
215 | smallest = NULL; |
216 | c = 0; | |
217 | ||
218 | for (j=0; j<cnt; j++) { | |
b414579f | 219 | rdev = conf->devlist[j]; |
0825b87a AN |
220 | printk(KERN_INFO "raid0: checking %s ...", |
221 | bdevname(rdev->bdev, b)); | |
d27a43ab | 222 | if (rdev->sectors <= zone->dev_start) { |
0825b87a | 223 | printk(KERN_INFO " nope.\n"); |
dd8ac336 AN |
224 | continue; |
225 | } | |
226 | printk(KERN_INFO " contained as device %d\n", c); | |
b414579f | 227 | dev[c] = rdev; |
dd8ac336 AN |
228 | c++; |
229 | if (!smallest || rdev->sectors < smallest->sectors) { | |
230 | smallest = rdev; | |
231 | printk(KERN_INFO " (%llu) is smallest!.\n", | |
232 | (unsigned long long)rdev->sectors); | |
233 | } | |
1da177e4 LT |
234 | } |
235 | ||
236 | zone->nb_dev = c; | |
49f357a2 | 237 | sectors = (smallest->sectors - zone->dev_start) * c; |
83838ed8 | 238 | printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n", |
49f357a2 | 239 | zone->nb_dev, (unsigned long long)sectors); |
1da177e4 | 240 | |
49f357a2 | 241 | curr_zone_end += sectors; |
d27a43ab | 242 | zone->zone_end = curr_zone_end; |
1da177e4 | 243 | |
6b8796cc | 244 | printk(KERN_INFO "raid0: current zone start: %llu\n", |
d27a43ab | 245 | (unsigned long long)smallest->sectors); |
1da177e4 | 246 | } |
1da177e4 | 247 | mddev->queue->unplug_fn = raid0_unplug; |
26be34dc N |
248 | mddev->queue->backing_dev_info.congested_fn = raid0_congested; |
249 | mddev->queue->backing_dev_info.congested_data = mddev; | |
1da177e4 | 250 | |
92e59b6b | 251 | /* |
252 | * now since we have the hard sector sizes, we can make sure | |
253 | * chunk size is a multiple of that sector size | |
254 | */ | |
9d8f0363 | 255 | if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) { |
92e59b6b | 256 | printk(KERN_ERR "%s chunk_size of %d not valid\n", |
257 | mdname(mddev), | |
9d8f0363 | 258 | mddev->chunk_sectors << 9); |
92e59b6b | 259 | goto abort; |
260 | } | |
8f6c2e4b MP |
261 | |
262 | blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); | |
263 | blk_queue_io_opt(mddev->queue, | |
264 | (mddev->chunk_sectors << 9) * mddev->raid_disks); | |
265 | ||
0825b87a | 266 | printk(KERN_INFO "raid0: done.\n"); |
ed7b0038 | 267 | mddev->private = conf; |
1da177e4 | 268 | return 0; |
5568a603 | 269 | abort: |
ed7b0038 AN |
270 | kfree(conf->strip_zone); |
271 | kfree(conf->devlist); | |
272 | kfree(conf); | |
273 | mddev->private = NULL; | |
274 | return err; | |
1da177e4 LT |
275 | } |
276 | ||
277 | /** | |
278 | * raid0_mergeable_bvec -- tell bio layer if a two requests can be merged | |
279 | * @q: request queue | |
cc371e66 | 280 | * @bvm: properties of new bio |
1da177e4 LT |
281 | * @biovec: the request that could be merged to it. |
282 | * | |
283 | * Return amount of bytes we can accept at this offset | |
284 | */ | |
cc371e66 AK |
285 | static int raid0_mergeable_bvec(struct request_queue *q, |
286 | struct bvec_merge_data *bvm, | |
287 | struct bio_vec *biovec) | |
1da177e4 LT |
288 | { |
289 | mddev_t *mddev = q->queuedata; | |
cc371e66 | 290 | sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev); |
1da177e4 | 291 | int max; |
9d8f0363 | 292 | unsigned int chunk_sectors = mddev->chunk_sectors; |
cc371e66 | 293 | unsigned int bio_sectors = bvm->bi_size >> 9; |
1da177e4 | 294 | |
d6e412ea | 295 | if (is_power_of_2(chunk_sectors)) |
fbb704ef | 296 | max = (chunk_sectors - ((sector & (chunk_sectors-1)) |
297 | + bio_sectors)) << 9; | |
298 | else | |
299 | max = (chunk_sectors - (sector_div(sector, chunk_sectors) | |
300 | + bio_sectors)) << 9; | |
1da177e4 LT |
301 | if (max < 0) max = 0; /* bio_add cannot handle a negative return */ |
302 | if (max <= biovec->bv_len && bio_sectors == 0) | |
303 | return biovec->bv_len; | |
304 | else | |
305 | return max; | |
306 | } | |
307 | ||
80c3a6ce DW |
308 | static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks) |
309 | { | |
310 | sector_t array_sectors = 0; | |
311 | mdk_rdev_t *rdev; | |
312 | ||
313 | WARN_ONCE(sectors || raid_disks, | |
314 | "%s does not support generic reshape\n", __func__); | |
315 | ||
316 | list_for_each_entry(rdev, &mddev->disks, same_set) | |
317 | array_sectors += rdev->sectors; | |
318 | ||
319 | return array_sectors; | |
320 | } | |
321 | ||
8f79cfcd | 322 | static int raid0_run(mddev_t *mddev) |
1da177e4 | 323 | { |
5568a603 | 324 | int ret; |
1da177e4 | 325 | |
9d8f0363 | 326 | if (mddev->chunk_sectors == 0) { |
fbb704ef | 327 | printk(KERN_ERR "md/raid0: chunk size must be set.\n"); |
2604b703 N |
328 | return -EINVAL; |
329 | } | |
0894cc30 AN |
330 | if (md_check_no_bitmap(mddev)) |
331 | return -EINVAL; | |
086fa5ff | 332 | blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); |
e7e72bf6 | 333 | mddev->queue->queue_lock = &mddev->queue->__queue_lock; |
1da177e4 | 334 | |
5568a603 AN |
335 | ret = create_strip_zones(mddev); |
336 | if (ret < 0) | |
ed7b0038 | 337 | return ret; |
1da177e4 LT |
338 | |
339 | /* calculate array device size */ | |
1f403624 | 340 | md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); |
1da177e4 | 341 | |
ccacc7d2 AN |
342 | printk(KERN_INFO "raid0 : md_size is %llu sectors.\n", |
343 | (unsigned long long)mddev->array_sectors); | |
1da177e4 LT |
344 | /* calculate the max read-ahead size. |
345 | * For read-ahead of large files to be effective, we need to | |
346 | * readahead at least twice a whole stripe. i.e. number of devices | |
347 | * multiplied by chunk size times 2. | |
348 | * If an individual device has an ra_pages greater than the | |
349 | * chunk size, then we will not drive that device as hard as it | |
350 | * wants. We consider this a configuration error: a larger | |
351 | * chunksize should be used in that case. | |
352 | */ | |
353 | { | |
9d8f0363 AN |
354 | int stripe = mddev->raid_disks * |
355 | (mddev->chunk_sectors << 9) / PAGE_SIZE; | |
1da177e4 LT |
356 | if (mddev->queue->backing_dev_info.ra_pages < 2* stripe) |
357 | mddev->queue->backing_dev_info.ra_pages = 2* stripe; | |
358 | } | |
359 | ||
1da177e4 | 360 | blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec); |
46994191 | 361 | dump_zones(mddev); |
ac5e7113 | 362 | md_integrity_register(mddev); |
1da177e4 | 363 | return 0; |
1da177e4 LT |
364 | } |
365 | ||
fb5ab4b5 | 366 | static int raid0_stop(mddev_t *mddev) |
1da177e4 | 367 | { |
070ec55d | 368 | raid0_conf_t *conf = mddev->private; |
1da177e4 LT |
369 | |
370 | blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ | |
990a8baf | 371 | kfree(conf->strip_zone); |
fb5ab4b5 | 372 | kfree(conf->devlist); |
990a8baf | 373 | kfree(conf); |
1da177e4 | 374 | mddev->private = NULL; |
1da177e4 LT |
375 | return 0; |
376 | } | |
377 | ||
49f357a2 N |
378 | /* Find the zone which holds a particular offset |
379 | * Update *sectorp to be an offset in that zone | |
380 | */ | |
dc582663 | 381 | static struct strip_zone *find_zone(struct raid0_private_data *conf, |
49f357a2 | 382 | sector_t *sectorp) |
dc582663 AN |
383 | { |
384 | int i; | |
385 | struct strip_zone *z = conf->strip_zone; | |
49f357a2 | 386 | sector_t sector = *sectorp; |
dc582663 AN |
387 | |
388 | for (i = 0; i < conf->nr_strip_zones; i++) | |
49f357a2 N |
389 | if (sector < z[i].zone_end) { |
390 | if (i) | |
391 | *sectorp = sector - z[i-1].zone_end; | |
dc582663 | 392 | return z + i; |
49f357a2 | 393 | } |
dc582663 AN |
394 | BUG(); |
395 | } | |
396 | ||
fbb704ef | 397 | /* |
398 | * remaps the bio to the target device. we separate two flows. | |
399 | * power 2 flow and a general flow for the sake of perfromance | |
400 | */ | |
401 | static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone, | |
402 | sector_t sector, sector_t *sector_offset) | |
1da177e4 | 403 | { |
fbb704ef | 404 | unsigned int sect_in_chunk; |
405 | sector_t chunk; | |
070ec55d | 406 | raid0_conf_t *conf = mddev->private; |
84707f38 | 407 | int raid_disks = conf->strip_zone[0].nb_dev; |
9d8f0363 | 408 | unsigned int chunk_sects = mddev->chunk_sectors; |
fbb704ef | 409 | |
d6e412ea | 410 | if (is_power_of_2(chunk_sects)) { |
fbb704ef | 411 | int chunksect_bits = ffz(~chunk_sects); |
412 | /* find the sector offset inside the chunk */ | |
413 | sect_in_chunk = sector & (chunk_sects - 1); | |
414 | sector >>= chunksect_bits; | |
415 | /* chunk in zone */ | |
416 | chunk = *sector_offset; | |
417 | /* quotient is the chunk in real device*/ | |
418 | sector_div(chunk, zone->nb_dev << chunksect_bits); | |
419 | } else{ | |
420 | sect_in_chunk = sector_div(sector, chunk_sects); | |
421 | chunk = *sector_offset; | |
422 | sector_div(chunk, chunk_sects * zone->nb_dev); | |
423 | } | |
424 | /* | |
425 | * position the bio over the real device | |
426 | * real sector = chunk in device + starting of zone | |
427 | * + the position in the chunk | |
428 | */ | |
429 | *sector_offset = (chunk * chunk_sects) + sect_in_chunk; | |
84707f38 | 430 | return conf->devlist[(zone - conf->strip_zone)*raid_disks |
fbb704ef | 431 | + sector_div(sector, zone->nb_dev)]; |
432 | } | |
433 | ||
434 | /* | |
435 | * Is io distribute over 1 or more chunks ? | |
436 | */ | |
437 | static inline int is_io_in_chunk_boundary(mddev_t *mddev, | |
438 | unsigned int chunk_sects, struct bio *bio) | |
439 | { | |
d6e412ea | 440 | if (likely(is_power_of_2(chunk_sects))) { |
fbb704ef | 441 | return chunk_sects >= ((bio->bi_sector & (chunk_sects-1)) |
442 | + (bio->bi_size >> 9)); | |
443 | } else{ | |
444 | sector_t sector = bio->bi_sector; | |
445 | return chunk_sects >= (sector_div(sector, chunk_sects) | |
446 | + (bio->bi_size >> 9)); | |
447 | } | |
448 | } | |
449 | ||
450 | static int raid0_make_request(struct request_queue *q, struct bio *bio) | |
451 | { | |
452 | mddev_t *mddev = q->queuedata; | |
453 | unsigned int chunk_sects; | |
454 | sector_t sector_offset; | |
1da177e4 LT |
455 | struct strip_zone *zone; |
456 | mdk_rdev_t *tmp_dev; | |
a362357b | 457 | const int rw = bio_data_dir(bio); |
c9959059 | 458 | int cpu; |
1da177e4 | 459 | |
1f98a13f | 460 | if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) { |
a2826aa9 | 461 | md_barrier_request(mddev, bio); |
e5dcdd80 N |
462 | return 0; |
463 | } | |
464 | ||
074a7aca TH |
465 | cpu = part_stat_lock(); |
466 | part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]); | |
467 | part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw], | |
468 | bio_sectors(bio)); | |
469 | part_stat_unlock(); | |
1da177e4 | 470 | |
9d8f0363 | 471 | chunk_sects = mddev->chunk_sectors; |
fbb704ef | 472 | if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) { |
473 | sector_t sector = bio->bi_sector; | |
1da177e4 LT |
474 | struct bio_pair *bp; |
475 | /* Sanity check -- queue functions should prevent this happening */ | |
476 | if (bio->bi_vcnt != 1 || | |
477 | bio->bi_idx != 0) | |
478 | goto bad_map; | |
479 | /* This is a one page bio that upper layers | |
480 | * refuse to split for us, so we need to split it. | |
481 | */ | |
d6e412ea | 482 | if (likely(is_power_of_2(chunk_sects))) |
fbb704ef | 483 | bp = bio_split(bio, chunk_sects - (sector & |
484 | (chunk_sects-1))); | |
485 | else | |
486 | bp = bio_split(bio, chunk_sects - | |
487 | sector_div(sector, chunk_sects)); | |
1da177e4 LT |
488 | if (raid0_make_request(q, &bp->bio1)) |
489 | generic_make_request(&bp->bio1); | |
490 | if (raid0_make_request(q, &bp->bio2)) | |
491 | generic_make_request(&bp->bio2); | |
492 | ||
493 | bio_pair_release(bp); | |
494 | return 0; | |
495 | } | |
1da177e4 | 496 | |
fbb704ef | 497 | sector_offset = bio->bi_sector; |
498 | zone = find_zone(mddev->private, §or_offset); | |
499 | tmp_dev = map_sector(mddev, zone, bio->bi_sector, | |
500 | §or_offset); | |
1da177e4 | 501 | bio->bi_bdev = tmp_dev->bdev; |
fbb704ef | 502 | bio->bi_sector = sector_offset + zone->dev_start + |
503 | tmp_dev->data_offset; | |
1da177e4 LT |
504 | /* |
505 | * Let the main block layer submit the IO and resolve recursion: | |
506 | */ | |
507 | return 1; | |
508 | ||
509 | bad_map: | |
510 | printk("raid0_make_request bug: can't convert block across chunks" | |
a4712005 | 511 | " or bigger than %dk %llu %d\n", chunk_sects / 2, |
1da177e4 LT |
512 | (unsigned long long)bio->bi_sector, bio->bi_size >> 10); |
513 | ||
6712ecf8 | 514 | bio_io_error(bio); |
1da177e4 LT |
515 | return 0; |
516 | } | |
8299d7f7 | 517 | |
1b961429 | 518 | static void raid0_status(struct seq_file *seq, mddev_t *mddev) |
1da177e4 LT |
519 | { |
520 | #undef MD_DEBUG | |
521 | #ifdef MD_DEBUG | |
522 | int j, k, h; | |
523 | char b[BDEVNAME_SIZE]; | |
070ec55d | 524 | raid0_conf_t *conf = mddev->private; |
84707f38 | 525 | int raid_disks = conf->strip_zone[0].nb_dev; |
8299d7f7 | 526 | |
1b961429 | 527 | sector_t zone_size; |
528 | sector_t zone_start = 0; | |
1da177e4 | 529 | h = 0; |
1b961429 | 530 | |
1da177e4 LT |
531 | for (j = 0; j < conf->nr_strip_zones; j++) { |
532 | seq_printf(seq, " z%d", j); | |
1da177e4 LT |
533 | seq_printf(seq, "=["); |
534 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) | |
8299d7f7 | 535 | seq_printf(seq, "%s/", bdevname( |
84707f38 | 536 | conf->devlist[j*raid_disks + k] |
1b961429 | 537 | ->bdev, b)); |
538 | ||
539 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
540 | seq_printf(seq, "] ze=%lld ds=%lld s=%lld\n", | |
541 | (unsigned long long)zone_start>>1, | |
542 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
543 | (unsigned long long)zone_size>>1); | |
544 | zone_start = conf->strip_zone[j].zone_end; | |
1da177e4 LT |
545 | } |
546 | #endif | |
9d8f0363 | 547 | seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); |
1da177e4 LT |
548 | return; |
549 | } | |
550 | ||
2604b703 | 551 | static struct mdk_personality raid0_personality= |
1da177e4 LT |
552 | { |
553 | .name = "raid0", | |
2604b703 | 554 | .level = 0, |
1da177e4 LT |
555 | .owner = THIS_MODULE, |
556 | .make_request = raid0_make_request, | |
557 | .run = raid0_run, | |
558 | .stop = raid0_stop, | |
559 | .status = raid0_status, | |
80c3a6ce | 560 | .size = raid0_size, |
1da177e4 LT |
561 | }; |
562 | ||
563 | static int __init raid0_init (void) | |
564 | { | |
2604b703 | 565 | return register_md_personality (&raid0_personality); |
1da177e4 LT |
566 | } |
567 | ||
568 | static void raid0_exit (void) | |
569 | { | |
2604b703 | 570 | unregister_md_personality (&raid0_personality); |
1da177e4 LT |
571 | } |
572 | ||
573 | module_init(raid0_init); | |
574 | module_exit(raid0_exit); | |
575 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 576 | MODULE_DESCRIPTION("RAID0 (striping) personality for MD"); |
1da177e4 | 577 | MODULE_ALIAS("md-personality-2"); /* RAID0 */ |
d9d166c2 | 578 | MODULE_ALIAS("md-raid0"); |
2604b703 | 579 | MODULE_ALIAS("md-level-0"); |