]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | raid0.c : Multiple Devices driver for Linux | |
f72ffdd6 | 3 | Copyright (C) 1994-96 Marc ZYNGIER |
1da177e4 LT |
4 | <zyngier@ufr-info-p7.ibp.fr> or |
5 | <maz@gloups.fdn.fr> | |
f72ffdd6 | 6 | Copyright (C) 1999, 2000 Ingo Molnar, Red Hat |
1da177e4 LT |
7 | |
8 | RAID-0 management functions. | |
9 | ||
10 | This program is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
f72ffdd6 | 14 | |
1da177e4 LT |
15 | You should have received a copy of the GNU General Public License |
16 | (for example /usr/src/linux/COPYING); if not, write to the Free | |
f72ffdd6 | 17 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
1da177e4 LT |
18 | */ |
19 | ||
bff61975 | 20 | #include <linux/blkdev.h> |
bff61975 | 21 | #include <linux/seq_file.h> |
056075c7 | 22 | #include <linux/module.h> |
5a0e3ad6 | 23 | #include <linux/slab.h> |
109e3765 | 24 | #include <trace/events/block.h> |
43b2e5d8 | 25 | #include "md.h" |
ef740c37 | 26 | #include "raid0.h" |
9af204cf | 27 | #include "raid5.h" |
1da177e4 | 28 | |
394ed8e4 SL |
29 | #define UNSUPPORTED_MDDEV_FLAGS \ |
30 | ((1L << MD_HAS_JOURNAL) | \ | |
31 | (1L << MD_JOURNAL_CLEAN) | \ | |
32 | (1L << MD_FAILFAST_SUPPORTED)) | |
33 | ||
5c675f83 | 34 | static int raid0_congested(struct mddev *mddev, int bits) |
26be34dc | 35 | { |
e373ab10 | 36 | struct r0conf *conf = mddev->private; |
3cb03002 | 37 | struct md_rdev **devlist = conf->devlist; |
84707f38 | 38 | int raid_disks = conf->strip_zone[0].nb_dev; |
26be34dc N |
39 | int i, ret = 0; |
40 | ||
84707f38 | 41 | for (i = 0; i < raid_disks && !ret ; i++) { |
165125e1 | 42 | struct request_queue *q = bdev_get_queue(devlist[i]->bdev); |
26be34dc | 43 | |
dc3b17cc | 44 | ret |= bdi_congested(q->backing_dev_info, bits); |
26be34dc N |
45 | } |
46 | return ret; | |
47 | } | |
48 | ||
46994191 | 49 | /* |
50 | * inform the user of the raid configuration | |
51 | */ | |
fd01b88c | 52 | static void dump_zones(struct mddev *mddev) |
46994191 | 53 | { |
50de8df4 | 54 | int j, k; |
46994191 | 55 | sector_t zone_size = 0; |
56 | sector_t zone_start = 0; | |
57 | char b[BDEVNAME_SIZE]; | |
e373ab10 | 58 | struct r0conf *conf = mddev->private; |
84707f38 | 59 | int raid_disks = conf->strip_zone[0].nb_dev; |
76603884 N |
60 | pr_debug("md: RAID0 configuration for %s - %d zone%s\n", |
61 | mdname(mddev), | |
62 | conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s"); | |
46994191 | 63 | for (j = 0; j < conf->nr_strip_zones; j++) { |
76603884 N |
64 | char line[200]; |
65 | int len = 0; | |
66 | ||
46994191 | 67 | for (k = 0; k < conf->strip_zone[j].nb_dev; k++) |
76603884 N |
68 | len += snprintf(line+len, 200-len, "%s%s", k?"/":"", |
69 | bdevname(conf->devlist[j*raid_disks | |
70 | + k]->bdev, b)); | |
71 | pr_debug("md: zone%d=[%s]\n", j, line); | |
46994191 | 72 | |
73 | zone_size = conf->strip_zone[j].zone_end - zone_start; | |
76603884 | 74 | pr_debug(" zone-offset=%10lluKB, device-offset=%10lluKB, size=%10lluKB\n", |
46994191 | 75 | (unsigned long long)zone_start>>1, |
76 | (unsigned long long)conf->strip_zone[j].dev_start>>1, | |
77 | (unsigned long long)zone_size>>1); | |
78 | zone_start = conf->strip_zone[j].zone_end; | |
79 | } | |
46994191 | 80 | } |
81 | ||
e373ab10 | 82 | static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf) |
1da177e4 | 83 | { |
a9f326eb | 84 | int i, c, err; |
49f357a2 | 85 | sector_t curr_zone_end, sectors; |
3cb03002 | 86 | struct md_rdev *smallest, *rdev1, *rdev2, *rdev, **dev; |
1da177e4 LT |
87 | struct strip_zone *zone; |
88 | int cnt; | |
89 | char b[BDEVNAME_SIZE]; | |
50de8df4 | 90 | char b2[BDEVNAME_SIZE]; |
e373ab10 | 91 | struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL); |
199dc6ed | 92 | unsigned short blksize = 512; |
ed7b0038 | 93 | |
7dedd15d | 94 | *private_conf = ERR_PTR(-ENOMEM); |
ed7b0038 AN |
95 | if (!conf) |
96 | return -ENOMEM; | |
dafb20fa | 97 | rdev_for_each(rdev1, mddev) { |
50de8df4 N |
98 | pr_debug("md/raid0:%s: looking at %s\n", |
99 | mdname(mddev), | |
100 | bdevname(rdev1->bdev, b)); | |
1da177e4 | 101 | c = 0; |
13f2682b N |
102 | |
103 | /* round size to chunk_size */ | |
104 | sectors = rdev1->sectors; | |
105 | sector_div(sectors, mddev->chunk_sectors); | |
106 | rdev1->sectors = sectors * mddev->chunk_sectors; | |
107 | ||
199dc6ed N |
108 | blksize = max(blksize, queue_logical_block_size( |
109 | rdev1->bdev->bd_disk->queue)); | |
110 | ||
dafb20fa | 111 | rdev_for_each(rdev2, mddev) { |
50de8df4 N |
112 | pr_debug("md/raid0:%s: comparing %s(%llu)" |
113 | " with %s(%llu)\n", | |
114 | mdname(mddev), | |
115 | bdevname(rdev1->bdev,b), | |
116 | (unsigned long long)rdev1->sectors, | |
117 | bdevname(rdev2->bdev,b2), | |
118 | (unsigned long long)rdev2->sectors); | |
1da177e4 | 119 | if (rdev2 == rdev1) { |
50de8df4 N |
120 | pr_debug("md/raid0:%s: END\n", |
121 | mdname(mddev)); | |
1da177e4 LT |
122 | break; |
123 | } | |
dd8ac336 | 124 | if (rdev2->sectors == rdev1->sectors) { |
1da177e4 LT |
125 | /* |
126 | * Not unique, don't count it as a new | |
127 | * group | |
128 | */ | |
50de8df4 N |
129 | pr_debug("md/raid0:%s: EQUAL\n", |
130 | mdname(mddev)); | |
1da177e4 LT |
131 | c = 1; |
132 | break; | |
133 | } | |
50de8df4 N |
134 | pr_debug("md/raid0:%s: NOT EQUAL\n", |
135 | mdname(mddev)); | |
1da177e4 LT |
136 | } |
137 | if (!c) { | |
50de8df4 N |
138 | pr_debug("md/raid0:%s: ==> UNIQUE\n", |
139 | mdname(mddev)); | |
1da177e4 | 140 | conf->nr_strip_zones++; |
50de8df4 N |
141 | pr_debug("md/raid0:%s: %d zones\n", |
142 | mdname(mddev), conf->nr_strip_zones); | |
1da177e4 LT |
143 | } |
144 | } | |
50de8df4 N |
145 | pr_debug("md/raid0:%s: FINAL %d zones\n", |
146 | mdname(mddev), conf->nr_strip_zones); | |
199dc6ed N |
147 | /* |
148 | * now since we have the hard sector sizes, we can make sure | |
149 | * chunk size is a multiple of that sector size | |
150 | */ | |
151 | if ((mddev->chunk_sectors << 9) % blksize) { | |
76603884 N |
152 | pr_warn("md/raid0:%s: chunk_size of %d not multiple of block size %d\n", |
153 | mdname(mddev), | |
154 | mddev->chunk_sectors << 9, blksize); | |
199dc6ed N |
155 | err = -EINVAL; |
156 | goto abort; | |
157 | } | |
158 | ||
ed7b0038 | 159 | err = -ENOMEM; |
9ffae0cf | 160 | conf->strip_zone = kzalloc(sizeof(struct strip_zone)* |
1da177e4 LT |
161 | conf->nr_strip_zones, GFP_KERNEL); |
162 | if (!conf->strip_zone) | |
ed7b0038 | 163 | goto abort; |
3cb03002 | 164 | conf->devlist = kzalloc(sizeof(struct md_rdev*)* |
1da177e4 LT |
165 | conf->nr_strip_zones*mddev->raid_disks, |
166 | GFP_KERNEL); | |
167 | if (!conf->devlist) | |
ed7b0038 | 168 | goto abort; |
1da177e4 | 169 | |
1da177e4 LT |
170 | /* The first zone must contain all devices, so here we check that |
171 | * there is a proper alignment of slots to devices and find them all | |
172 | */ | |
173 | zone = &conf->strip_zone[0]; | |
174 | cnt = 0; | |
175 | smallest = NULL; | |
b414579f | 176 | dev = conf->devlist; |
ed7b0038 | 177 | err = -EINVAL; |
dafb20fa | 178 | rdev_for_each(rdev1, mddev) { |
1da177e4 LT |
179 | int j = rdev1->raid_disk; |
180 | ||
e93f68a1 | 181 | if (mddev->level == 10) { |
9af204cf TM |
182 | /* taking over a raid10-n2 array */ |
183 | j /= 2; | |
e93f68a1 N |
184 | rdev1->new_raid_disk = j; |
185 | } | |
9af204cf | 186 | |
fc3a08b8 KW |
187 | if (mddev->level == 1) { |
188 | /* taiking over a raid1 array- | |
189 | * we have only one active disk | |
190 | */ | |
191 | j = 0; | |
192 | rdev1->new_raid_disk = j; | |
193 | } | |
194 | ||
f96c9f30 | 195 | if (j < 0) { |
76603884 N |
196 | pr_warn("md/raid0:%s: remove inactive devices before converting to RAID0\n", |
197 | mdname(mddev)); | |
f96c9f30 N |
198 | goto abort; |
199 | } | |
200 | if (j >= mddev->raid_disks) { | |
76603884 N |
201 | pr_warn("md/raid0:%s: bad disk number %d - aborting!\n", |
202 | mdname(mddev), j); | |
1da177e4 LT |
203 | goto abort; |
204 | } | |
b414579f | 205 | if (dev[j]) { |
76603884 N |
206 | pr_warn("md/raid0:%s: multiple devices for %d - aborting!\n", |
207 | mdname(mddev), j); | |
1da177e4 LT |
208 | goto abort; |
209 | } | |
b414579f | 210 | dev[j] = rdev1; |
1da177e4 | 211 | |
dd8ac336 | 212 | if (!smallest || (rdev1->sectors < smallest->sectors)) |
1da177e4 LT |
213 | smallest = rdev1; |
214 | cnt++; | |
215 | } | |
216 | if (cnt != mddev->raid_disks) { | |
76603884 N |
217 | pr_warn("md/raid0:%s: too few disks (%d of %d) - aborting!\n", |
218 | mdname(mddev), cnt, mddev->raid_disks); | |
1da177e4 LT |
219 | goto abort; |
220 | } | |
221 | zone->nb_dev = cnt; | |
49f357a2 | 222 | zone->zone_end = smallest->sectors * cnt; |
1da177e4 | 223 | |
49f357a2 | 224 | curr_zone_end = zone->zone_end; |
1da177e4 LT |
225 | |
226 | /* now do the other zones */ | |
227 | for (i = 1; i < conf->nr_strip_zones; i++) | |
228 | { | |
a9f326eb N |
229 | int j; |
230 | ||
1da177e4 | 231 | zone = conf->strip_zone + i; |
b414579f | 232 | dev = conf->devlist + i * mddev->raid_disks; |
1da177e4 | 233 | |
50de8df4 | 234 | pr_debug("md/raid0:%s: zone %d\n", mdname(mddev), i); |
d27a43ab | 235 | zone->dev_start = smallest->sectors; |
1da177e4 LT |
236 | smallest = NULL; |
237 | c = 0; | |
238 | ||
239 | for (j=0; j<cnt; j++) { | |
b414579f | 240 | rdev = conf->devlist[j]; |
d27a43ab | 241 | if (rdev->sectors <= zone->dev_start) { |
50de8df4 N |
242 | pr_debug("md/raid0:%s: checking %s ... nope\n", |
243 | mdname(mddev), | |
244 | bdevname(rdev->bdev, b)); | |
dd8ac336 AN |
245 | continue; |
246 | } | |
50de8df4 N |
247 | pr_debug("md/raid0:%s: checking %s ..." |
248 | " contained as device %d\n", | |
249 | mdname(mddev), | |
250 | bdevname(rdev->bdev, b), c); | |
b414579f | 251 | dev[c] = rdev; |
dd8ac336 AN |
252 | c++; |
253 | if (!smallest || rdev->sectors < smallest->sectors) { | |
254 | smallest = rdev; | |
50de8df4 N |
255 | pr_debug("md/raid0:%s: (%llu) is smallest!.\n", |
256 | mdname(mddev), | |
257 | (unsigned long long)rdev->sectors); | |
dd8ac336 | 258 | } |
1da177e4 LT |
259 | } |
260 | ||
261 | zone->nb_dev = c; | |
49f357a2 | 262 | sectors = (smallest->sectors - zone->dev_start) * c; |
50de8df4 N |
263 | pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n", |
264 | mdname(mddev), | |
265 | zone->nb_dev, (unsigned long long)sectors); | |
1da177e4 | 266 | |
49f357a2 | 267 | curr_zone_end += sectors; |
d27a43ab | 268 | zone->zone_end = curr_zone_end; |
1da177e4 | 269 | |
50de8df4 N |
270 | pr_debug("md/raid0:%s: current zone start: %llu\n", |
271 | mdname(mddev), | |
272 | (unsigned long long)smallest->sectors); | |
1da177e4 | 273 | } |
1da177e4 | 274 | |
50de8df4 | 275 | pr_debug("md/raid0:%s: done.\n", mdname(mddev)); |
9af204cf TM |
276 | *private_conf = conf; |
277 | ||
1da177e4 | 278 | return 0; |
5568a603 | 279 | abort: |
ed7b0038 AN |
280 | kfree(conf->strip_zone); |
281 | kfree(conf->devlist); | |
282 | kfree(conf); | |
58ebb34c | 283 | *private_conf = ERR_PTR(err); |
ed7b0038 | 284 | return err; |
1da177e4 LT |
285 | } |
286 | ||
ba13da47 N |
287 | /* Find the zone which holds a particular offset |
288 | * Update *sectorp to be an offset in that zone | |
289 | */ | |
290 | static struct strip_zone *find_zone(struct r0conf *conf, | |
291 | sector_t *sectorp) | |
292 | { | |
293 | int i; | |
294 | struct strip_zone *z = conf->strip_zone; | |
295 | sector_t sector = *sectorp; | |
296 | ||
297 | for (i = 0; i < conf->nr_strip_zones; i++) | |
298 | if (sector < z[i].zone_end) { | |
299 | if (i) | |
300 | *sectorp = sector - z[i-1].zone_end; | |
301 | return z + i; | |
302 | } | |
303 | BUG(); | |
304 | } | |
305 | ||
306 | /* | |
307 | * remaps the bio to the target device. we separate two flows. | |
47d68979 | 308 | * power 2 flow and a general flow for the sake of performance |
ba13da47 N |
309 | */ |
310 | static struct md_rdev *map_sector(struct mddev *mddev, struct strip_zone *zone, | |
311 | sector_t sector, sector_t *sector_offset) | |
312 | { | |
313 | unsigned int sect_in_chunk; | |
314 | sector_t chunk; | |
315 | struct r0conf *conf = mddev->private; | |
316 | int raid_disks = conf->strip_zone[0].nb_dev; | |
317 | unsigned int chunk_sects = mddev->chunk_sectors; | |
318 | ||
319 | if (is_power_of_2(chunk_sects)) { | |
320 | int chunksect_bits = ffz(~chunk_sects); | |
321 | /* find the sector offset inside the chunk */ | |
322 | sect_in_chunk = sector & (chunk_sects - 1); | |
323 | sector >>= chunksect_bits; | |
324 | /* chunk in zone */ | |
325 | chunk = *sector_offset; | |
326 | /* quotient is the chunk in real device*/ | |
327 | sector_div(chunk, zone->nb_dev << chunksect_bits); | |
328 | } else{ | |
329 | sect_in_chunk = sector_div(sector, chunk_sects); | |
330 | chunk = *sector_offset; | |
331 | sector_div(chunk, chunk_sects * zone->nb_dev); | |
332 | } | |
333 | /* | |
334 | * position the bio over the real device | |
335 | * real sector = chunk in device + starting of zone | |
336 | * + the position in the chunk | |
337 | */ | |
338 | *sector_offset = (chunk * chunk_sects) + sect_in_chunk; | |
339 | return conf->devlist[(zone - conf->strip_zone)*raid_disks | |
340 | + sector_div(sector, zone->nb_dev)]; | |
341 | } | |
342 | ||
fd01b88c | 343 | static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks) |
80c3a6ce DW |
344 | { |
345 | sector_t array_sectors = 0; | |
3cb03002 | 346 | struct md_rdev *rdev; |
80c3a6ce DW |
347 | |
348 | WARN_ONCE(sectors || raid_disks, | |
349 | "%s does not support generic reshape\n", __func__); | |
350 | ||
dafb20fa | 351 | rdev_for_each(rdev, mddev) |
a6468539 N |
352 | array_sectors += (rdev->sectors & |
353 | ~(sector_t)(mddev->chunk_sectors-1)); | |
80c3a6ce DW |
354 | |
355 | return array_sectors; | |
356 | } | |
357 | ||
afa0f557 | 358 | static void raid0_free(struct mddev *mddev, void *priv); |
0366ef84 | 359 | |
fd01b88c | 360 | static int raid0_run(struct mddev *mddev) |
1da177e4 | 361 | { |
e373ab10 | 362 | struct r0conf *conf; |
5568a603 | 363 | int ret; |
1da177e4 | 364 | |
9d8f0363 | 365 | if (mddev->chunk_sectors == 0) { |
76603884 | 366 | pr_warn("md/raid0:%s: chunk size must be set.\n", mdname(mddev)); |
2604b703 N |
367 | return -EINVAL; |
368 | } | |
0894cc30 AN |
369 | if (md_check_no_bitmap(mddev)) |
370 | return -EINVAL; | |
753f2856 | 371 | |
9af204cf TM |
372 | /* if private is not null, we are here after takeover */ |
373 | if (mddev->private == NULL) { | |
374 | ret = create_strip_zones(mddev, &conf); | |
375 | if (ret < 0) | |
376 | return ret; | |
377 | mddev->private = conf; | |
378 | } | |
379 | conf = mddev->private; | |
199dc6ed N |
380 | if (mddev->queue) { |
381 | struct md_rdev *rdev; | |
382 | bool discard_supported = false; | |
383 | ||
199dc6ed N |
384 | blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); |
385 | blk_queue_max_write_same_sectors(mddev->queue, mddev->chunk_sectors); | |
386 | blk_queue_max_discard_sectors(mddev->queue, mddev->chunk_sectors); | |
387 | ||
388 | blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); | |
389 | blk_queue_io_opt(mddev->queue, | |
390 | (mddev->chunk_sectors << 9) * mddev->raid_disks); | |
391 | ||
66eefe5d N |
392 | rdev_for_each(rdev, mddev) { |
393 | disk_stack_limits(mddev->gendisk, rdev->bdev, | |
394 | rdev->data_offset << 9); | |
395 | if (blk_queue_discard(bdev_get_queue(rdev->bdev))) | |
396 | discard_supported = true; | |
397 | } | |
199dc6ed N |
398 | if (!discard_supported) |
399 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue); | |
400 | else | |
401 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue); | |
402 | } | |
1da177e4 LT |
403 | |
404 | /* calculate array device size */ | |
1f403624 | 405 | md_set_array_sectors(mddev, raid0_size(mddev, 0, 0)); |
1da177e4 | 406 | |
76603884 N |
407 | pr_debug("md/raid0:%s: md_size is %llu sectors.\n", |
408 | mdname(mddev), | |
409 | (unsigned long long)mddev->array_sectors); | |
753f2856 HM |
410 | |
411 | if (mddev->queue) { | |
412 | /* calculate the max read-ahead size. | |
413 | * For read-ahead of large files to be effective, we need to | |
414 | * readahead at least twice a whole stripe. i.e. number of devices | |
415 | * multiplied by chunk size times 2. | |
416 | * If an individual device has an ra_pages greater than the | |
417 | * chunk size, then we will not drive that device as hard as it | |
418 | * wants. We consider this a configuration error: a larger | |
419 | * chunksize should be used in that case. | |
420 | */ | |
9d8f0363 AN |
421 | int stripe = mddev->raid_disks * |
422 | (mddev->chunk_sectors << 9) / PAGE_SIZE; | |
dc3b17cc JK |
423 | if (mddev->queue->backing_dev_info->ra_pages < 2* stripe) |
424 | mddev->queue->backing_dev_info->ra_pages = 2* stripe; | |
1da177e4 LT |
425 | } |
426 | ||
46994191 | 427 | dump_zones(mddev); |
0366ef84 | 428 | |
429 | ret = md_integrity_register(mddev); | |
0366ef84 | 430 | |
431 | return ret; | |
1da177e4 LT |
432 | } |
433 | ||
afa0f557 | 434 | static void raid0_free(struct mddev *mddev, void *priv) |
1da177e4 | 435 | { |
afa0f557 | 436 | struct r0conf *conf = priv; |
1da177e4 | 437 | |
990a8baf | 438 | kfree(conf->strip_zone); |
fb5ab4b5 | 439 | kfree(conf->devlist); |
990a8baf | 440 | kfree(conf); |
1da177e4 LT |
441 | } |
442 | ||
fbb704ef | 443 | /* |
444 | * Is io distribute over 1 or more chunks ? | |
445 | */ | |
fd01b88c | 446 | static inline int is_io_in_chunk_boundary(struct mddev *mddev, |
fbb704ef | 447 | unsigned int chunk_sects, struct bio *bio) |
448 | { | |
d6e412ea | 449 | if (likely(is_power_of_2(chunk_sects))) { |
4f024f37 KO |
450 | return chunk_sects >= |
451 | ((bio->bi_iter.bi_sector & (chunk_sects-1)) | |
aa8b57aa | 452 | + bio_sectors(bio)); |
fbb704ef | 453 | } else{ |
4f024f37 | 454 | sector_t sector = bio->bi_iter.bi_sector; |
fbb704ef | 455 | return chunk_sects >= (sector_div(sector, chunk_sects) |
aa8b57aa | 456 | + bio_sectors(bio)); |
fbb704ef | 457 | } |
458 | } | |
459 | ||
b4fdcb02 | 460 | static void raid0_make_request(struct mddev *mddev, struct bio *bio) |
fbb704ef | 461 | { |
1da177e4 | 462 | struct strip_zone *zone; |
3cb03002 | 463 | struct md_rdev *tmp_dev; |
20d0189b | 464 | struct bio *split; |
1da177e4 | 465 | |
1eff9d32 | 466 | if (unlikely(bio->bi_opf & REQ_PREFLUSH)) { |
e9c7469b | 467 | md_flush_request(mddev, bio); |
5a7bbad2 | 468 | return; |
e5dcdd80 N |
469 | } |
470 | ||
20d0189b | 471 | do { |
109e3765 N |
472 | sector_t bio_sector = bio->bi_iter.bi_sector; |
473 | sector_t sector = bio_sector; | |
20d0189b KO |
474 | unsigned chunk_sects = mddev->chunk_sectors; |
475 | ||
476 | unsigned sectors = chunk_sects - | |
477 | (likely(is_power_of_2(chunk_sects)) | |
478 | ? (sector & (chunk_sects-1)) | |
479 | : sector_div(sector, chunk_sects)); | |
480 | ||
a8115776 | 481 | /* Restore due to sector_div */ |
109e3765 | 482 | sector = bio_sector; |
a8115776 | 483 | |
20d0189b KO |
484 | if (sectors < bio_sectors(bio)) { |
485 | split = bio_split(bio, sectors, GFP_NOIO, fs_bio_set); | |
486 | bio_chain(split, bio); | |
487 | } else { | |
488 | split = bio; | |
489 | } | |
1da177e4 | 490 | |
20d0189b KO |
491 | zone = find_zone(mddev->private, §or); |
492 | tmp_dev = map_sector(mddev, zone, sector, §or); | |
493 | split->bi_bdev = tmp_dev->bdev; | |
494 | split->bi_iter.bi_sector = sector + zone->dev_start + | |
495 | tmp_dev->data_offset; | |
496 | ||
796a5cf0 | 497 | if (unlikely((bio_op(split) == REQ_OP_DISCARD) && |
20d0189b KO |
498 | !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) { |
499 | /* Just ignore it */ | |
4246a0b6 | 500 | bio_endio(split); |
109e3765 N |
501 | } else { |
502 | if (mddev->gendisk) | |
503 | trace_block_bio_remap(bdev_get_queue(split->bi_bdev), | |
504 | split, disk_devt(mddev->gendisk), | |
505 | bio_sector); | |
26483819 | 506 | mddev_check_writesame(mddev, split); |
20d0189b | 507 | generic_make_request(split); |
109e3765 | 508 | } |
20d0189b | 509 | } while (split != bio); |
1da177e4 | 510 | } |
8299d7f7 | 511 | |
fd01b88c | 512 | static void raid0_status(struct seq_file *seq, struct mddev *mddev) |
1da177e4 | 513 | { |
9d8f0363 | 514 | seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2); |
1da177e4 LT |
515 | return; |
516 | } | |
517 | ||
fd01b88c | 518 | static void *raid0_takeover_raid45(struct mddev *mddev) |
9af204cf | 519 | { |
3cb03002 | 520 | struct md_rdev *rdev; |
e373ab10 | 521 | struct r0conf *priv_conf; |
9af204cf TM |
522 | |
523 | if (mddev->degraded != 1) { | |
76603884 N |
524 | pr_warn("md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n", |
525 | mdname(mddev), | |
526 | mddev->degraded); | |
9af204cf TM |
527 | return ERR_PTR(-EINVAL); |
528 | } | |
529 | ||
dafb20fa | 530 | rdev_for_each(rdev, mddev) { |
9af204cf TM |
531 | /* check slot number for a disk */ |
532 | if (rdev->raid_disk == mddev->raid_disks-1) { | |
76603884 N |
533 | pr_warn("md/raid0:%s: raid5 must have missing parity disk!\n", |
534 | mdname(mddev)); | |
9af204cf TM |
535 | return ERR_PTR(-EINVAL); |
536 | } | |
eea136d6 | 537 | rdev->sectors = mddev->dev_sectors; |
9af204cf TM |
538 | } |
539 | ||
540 | /* Set new parameters */ | |
541 | mddev->new_level = 0; | |
001048a3 | 542 | mddev->new_layout = 0; |
9af204cf TM |
543 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
544 | mddev->raid_disks--; | |
545 | mddev->delta_disks = -1; | |
546 | /* make sure it will be not marked as dirty */ | |
547 | mddev->recovery_cp = MaxSector; | |
394ed8e4 | 548 | mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS); |
9af204cf TM |
549 | |
550 | create_strip_zones(mddev, &priv_conf); | |
6995f0b2 | 551 | |
9af204cf TM |
552 | return priv_conf; |
553 | } | |
554 | ||
fd01b88c | 555 | static void *raid0_takeover_raid10(struct mddev *mddev) |
9af204cf | 556 | { |
e373ab10 | 557 | struct r0conf *priv_conf; |
9af204cf TM |
558 | |
559 | /* Check layout: | |
560 | * - far_copies must be 1 | |
561 | * - near_copies must be 2 | |
562 | * - disks number must be even | |
563 | * - all mirrors must be already degraded | |
564 | */ | |
565 | if (mddev->layout != ((1 << 8) + 2)) { | |
76603884 N |
566 | pr_warn("md/raid0:%s:: Raid0 cannot takeover layout: 0x%x\n", |
567 | mdname(mddev), | |
568 | mddev->layout); | |
9af204cf TM |
569 | return ERR_PTR(-EINVAL); |
570 | } | |
571 | if (mddev->raid_disks & 1) { | |
76603884 N |
572 | pr_warn("md/raid0:%s: Raid0 cannot takeover Raid10 with odd disk number.\n", |
573 | mdname(mddev)); | |
9af204cf TM |
574 | return ERR_PTR(-EINVAL); |
575 | } | |
576 | if (mddev->degraded != (mddev->raid_disks>>1)) { | |
76603884 N |
577 | pr_warn("md/raid0:%s: All mirrors must be already degraded!\n", |
578 | mdname(mddev)); | |
9af204cf TM |
579 | return ERR_PTR(-EINVAL); |
580 | } | |
581 | ||
582 | /* Set new parameters */ | |
583 | mddev->new_level = 0; | |
001048a3 | 584 | mddev->new_layout = 0; |
9af204cf TM |
585 | mddev->new_chunk_sectors = mddev->chunk_sectors; |
586 | mddev->delta_disks = - mddev->raid_disks / 2; | |
587 | mddev->raid_disks += mddev->delta_disks; | |
588 | mddev->degraded = 0; | |
589 | /* make sure it will be not marked as dirty */ | |
590 | mddev->recovery_cp = MaxSector; | |
394ed8e4 | 591 | mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS); |
9af204cf TM |
592 | |
593 | create_strip_zones(mddev, &priv_conf); | |
9af204cf TM |
594 | return priv_conf; |
595 | } | |
596 | ||
fd01b88c | 597 | static void *raid0_takeover_raid1(struct mddev *mddev) |
fc3a08b8 | 598 | { |
e373ab10 | 599 | struct r0conf *priv_conf; |
24b961f8 | 600 | int chunksect; |
fc3a08b8 KW |
601 | |
602 | /* Check layout: | |
603 | * - (N - 1) mirror drives must be already faulty | |
604 | */ | |
605 | if ((mddev->raid_disks - 1) != mddev->degraded) { | |
76603884 | 606 | pr_err("md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n", |
fc3a08b8 KW |
607 | mdname(mddev)); |
608 | return ERR_PTR(-EINVAL); | |
609 | } | |
610 | ||
24b961f8 JS |
611 | /* |
612 | * a raid1 doesn't have the notion of chunk size, so | |
613 | * figure out the largest suitable size we can use. | |
614 | */ | |
615 | chunksect = 64 * 2; /* 64K by default */ | |
616 | ||
617 | /* The array must be an exact multiple of chunksize */ | |
618 | while (chunksect && (mddev->array_sectors & (chunksect - 1))) | |
619 | chunksect >>= 1; | |
620 | ||
621 | if ((chunksect << 9) < PAGE_SIZE) | |
622 | /* array size does not allow a suitable chunk size */ | |
623 | return ERR_PTR(-EINVAL); | |
624 | ||
fc3a08b8 KW |
625 | /* Set new parameters */ |
626 | mddev->new_level = 0; | |
627 | mddev->new_layout = 0; | |
24b961f8 JS |
628 | mddev->new_chunk_sectors = chunksect; |
629 | mddev->chunk_sectors = chunksect; | |
fc3a08b8 | 630 | mddev->delta_disks = 1 - mddev->raid_disks; |
f7bee809 | 631 | mddev->raid_disks = 1; |
fc3a08b8 KW |
632 | /* make sure it will be not marked as dirty */ |
633 | mddev->recovery_cp = MaxSector; | |
394ed8e4 | 634 | mddev_clear_unsupported_flags(mddev, UNSUPPORTED_MDDEV_FLAGS); |
fc3a08b8 KW |
635 | |
636 | create_strip_zones(mddev, &priv_conf); | |
637 | return priv_conf; | |
638 | } | |
639 | ||
fd01b88c | 640 | static void *raid0_takeover(struct mddev *mddev) |
9af204cf TM |
641 | { |
642 | /* raid0 can take over: | |
049d6c1e | 643 | * raid4 - if all data disks are active. |
9af204cf TM |
644 | * raid5 - providing it is Raid4 layout and one disk is faulty |
645 | * raid10 - assuming we have all necessary active disks | |
fc3a08b8 | 646 | * raid1 - with (N -1) mirror drives faulty |
9af204cf | 647 | */ |
a8461a61 N |
648 | |
649 | if (mddev->bitmap) { | |
76603884 N |
650 | pr_warn("md/raid0: %s: cannot takeover array with bitmap\n", |
651 | mdname(mddev)); | |
a8461a61 N |
652 | return ERR_PTR(-EBUSY); |
653 | } | |
049d6c1e MT |
654 | if (mddev->level == 4) |
655 | return raid0_takeover_raid45(mddev); | |
656 | ||
9af204cf TM |
657 | if (mddev->level == 5) { |
658 | if (mddev->layout == ALGORITHM_PARITY_N) | |
049d6c1e | 659 | return raid0_takeover_raid45(mddev); |
9af204cf | 660 | |
76603884 N |
661 | pr_warn("md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n", |
662 | mdname(mddev), ALGORITHM_PARITY_N); | |
9af204cf TM |
663 | } |
664 | ||
665 | if (mddev->level == 10) | |
666 | return raid0_takeover_raid10(mddev); | |
667 | ||
fc3a08b8 KW |
668 | if (mddev->level == 1) |
669 | return raid0_takeover_raid1(mddev); | |
670 | ||
76603884 | 671 | pr_warn("Takeover from raid%i to raid0 not supported\n", |
fc3a08b8 KW |
672 | mddev->level); |
673 | ||
9af204cf TM |
674 | return ERR_PTR(-EINVAL); |
675 | } | |
676 | ||
fd01b88c | 677 | static void raid0_quiesce(struct mddev *mddev, int state) |
9af204cf TM |
678 | { |
679 | } | |
680 | ||
84fc4b56 | 681 | static struct md_personality raid0_personality= |
1da177e4 LT |
682 | { |
683 | .name = "raid0", | |
2604b703 | 684 | .level = 0, |
1da177e4 LT |
685 | .owner = THIS_MODULE, |
686 | .make_request = raid0_make_request, | |
687 | .run = raid0_run, | |
afa0f557 | 688 | .free = raid0_free, |
1da177e4 | 689 | .status = raid0_status, |
80c3a6ce | 690 | .size = raid0_size, |
9af204cf TM |
691 | .takeover = raid0_takeover, |
692 | .quiesce = raid0_quiesce, | |
5c675f83 | 693 | .congested = raid0_congested, |
1da177e4 LT |
694 | }; |
695 | ||
696 | static int __init raid0_init (void) | |
697 | { | |
2604b703 | 698 | return register_md_personality (&raid0_personality); |
1da177e4 LT |
699 | } |
700 | ||
701 | static void raid0_exit (void) | |
702 | { | |
2604b703 | 703 | unregister_md_personality (&raid0_personality); |
1da177e4 LT |
704 | } |
705 | ||
706 | module_init(raid0_init); | |
707 | module_exit(raid0_exit); | |
708 | MODULE_LICENSE("GPL"); | |
0efb9e61 | 709 | MODULE_DESCRIPTION("RAID0 (striping) personality for MD"); |
1da177e4 | 710 | MODULE_ALIAS("md-personality-2"); /* RAID0 */ |
d9d166c2 | 711 | MODULE_ALIAS("md-raid0"); |
2604b703 | 712 | MODULE_ALIAS("md-level-0"); |