raid0.c 18.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
   raid0.c : Multiple Devices driver for Linux
             Copyright (C) 1994-96 Marc ZYNGIER
	     <zyngier@ufr-info-p7.ibp.fr> or
	     <maz@gloups.fdn.fr>
             Copyright (C) 1999, 2000 Ingo Molnar, Red Hat


   RAID-0 management functions.

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.
   
   You should have received a copy of the GNU General Public License
   (for example /usr/src/linux/COPYING); if not, write to the Free
   Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  
*/

21 22
#include <linux/blkdev.h>
#include <linux/seq_file.h>
23
#include <linux/slab.h>
24
#include "md.h"
25
#include "raid0.h"
26
#include "raid5.h"
L
Linus Torvalds 已提交
27

28 29 30
static int raid0_congested(void *data, int bits)
{
	mddev_t *mddev = data;
31
	raid0_conf_t *conf = mddev->private;
32
	mdk_rdev_t **devlist = conf->devlist;
33
	int raid_disks = conf->strip_zone[0].nb_dev;
34 35
	int i, ret = 0;

36 37 38
	if (mddev_congested(mddev, bits))
		return 1;

39
	for (i = 0; i < raid_disks && !ret ; i++) {
40
		struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
41 42 43 44 45 46

		ret |= bdi_congested(&q->backing_dev_info, bits);
	}
	return ret;
}

47 48 49 50 51
/*
 * inform the user of the raid configuration
*/
static void dump_zones(mddev_t *mddev)
{
52
	int j, k;
53 54 55 56
	sector_t zone_size = 0;
	sector_t zone_start = 0;
	char b[BDEVNAME_SIZE];
	raid0_conf_t *conf = mddev->private;
57
	int raid_disks = conf->strip_zone[0].nb_dev;
58 59 60
	printk(KERN_INFO "md: RAID0 configuration for %s - %d zone%s\n",
	       mdname(mddev),
	       conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
61
	for (j = 0; j < conf->nr_strip_zones; j++) {
62
		printk(KERN_INFO "md: zone%d=[", j);
63
		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
64
			printk(KERN_CONT "%s%s", k?"/":"",
65
			bdevname(conf->devlist[j*raid_disks
66
						+ k]->bdev, b));
N
NeilBrown 已提交
67
		printk(KERN_CONT "]\n");
68 69

		zone_size  = conf->strip_zone[j].zone_end - zone_start;
70 71
		printk(KERN_INFO "      zone-offset=%10lluKB, "
				"device-offset=%10lluKB, size=%10lluKB\n",
72 73 74 75 76
			(unsigned long long)zone_start>>1,
			(unsigned long long)conf->strip_zone[j].dev_start>>1,
			(unsigned long long)zone_size>>1);
		zone_start = conf->strip_zone[j].zone_end;
	}
77
	printk(KERN_INFO "\n");
78 79
}

80
static int create_strip_zones(mddev_t *mddev, raid0_conf_t **private_conf)
L
Linus Torvalds 已提交
81
{
82
	int i, c, err;
83
	sector_t curr_zone_end, sectors;
84
	mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev, **dev;
L
Linus Torvalds 已提交
85 86 87
	struct strip_zone *zone;
	int cnt;
	char b[BDEVNAME_SIZE];
88
	char b2[BDEVNAME_SIZE];
89 90 91 92
	raid0_conf_t *conf = kzalloc(sizeof(*conf), GFP_KERNEL);

	if (!conf)
		return -ENOMEM;
93
	list_for_each_entry(rdev1, &mddev->disks, same_set) {
94 95 96
		pr_debug("md/raid0:%s: looking at %s\n",
			 mdname(mddev),
			 bdevname(rdev1->bdev, b));
L
Linus Torvalds 已提交
97
		c = 0;
98 99 100 101 102 103

		/* round size to chunk_size */
		sectors = rdev1->sectors;
		sector_div(sectors, mddev->chunk_sectors);
		rdev1->sectors = sectors * mddev->chunk_sectors;

104
		list_for_each_entry(rdev2, &mddev->disks, same_set) {
105 106 107 108 109 110 111
			pr_debug("md/raid0:%s:   comparing %s(%llu)"
				 " with %s(%llu)\n",
				 mdname(mddev),
				 bdevname(rdev1->bdev,b),
				 (unsigned long long)rdev1->sectors,
				 bdevname(rdev2->bdev,b2),
				 (unsigned long long)rdev2->sectors);
L
Linus Torvalds 已提交
112
			if (rdev2 == rdev1) {
113 114
				pr_debug("md/raid0:%s:   END\n",
					 mdname(mddev));
L
Linus Torvalds 已提交
115 116
				break;
			}
117
			if (rdev2->sectors == rdev1->sectors) {
L
Linus Torvalds 已提交
118 119 120 121
				/*
				 * Not unique, don't count it as a new
				 * group
				 */
122 123
				pr_debug("md/raid0:%s:   EQUAL\n",
					 mdname(mddev));
L
Linus Torvalds 已提交
124 125 126
				c = 1;
				break;
			}
127 128
			pr_debug("md/raid0:%s:   NOT EQUAL\n",
				 mdname(mddev));
L
Linus Torvalds 已提交
129 130
		}
		if (!c) {
131 132
			pr_debug("md/raid0:%s:   ==> UNIQUE\n",
				 mdname(mddev));
L
Linus Torvalds 已提交
133
			conf->nr_strip_zones++;
134 135
			pr_debug("md/raid0:%s: %d zones\n",
				 mdname(mddev), conf->nr_strip_zones);
L
Linus Torvalds 已提交
136 137
		}
	}
138 139
	pr_debug("md/raid0:%s: FINAL %d zones\n",
		 mdname(mddev), conf->nr_strip_zones);
140
	err = -ENOMEM;
141
	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
L
Linus Torvalds 已提交
142 143
				conf->nr_strip_zones, GFP_KERNEL);
	if (!conf->strip_zone)
144
		goto abort;
145
	conf->devlist = kzalloc(sizeof(mdk_rdev_t*)*
L
Linus Torvalds 已提交
146 147 148
				conf->nr_strip_zones*mddev->raid_disks,
				GFP_KERNEL);
	if (!conf->devlist)
149
		goto abort;
L
Linus Torvalds 已提交
150 151 152 153 154 155 156

	/* The first zone must contain all devices, so here we check that
	 * there is a proper alignment of slots to devices and find them all
	 */
	zone = &conf->strip_zone[0];
	cnt = 0;
	smallest = NULL;
157
	dev = conf->devlist;
158
	err = -EINVAL;
159
	list_for_each_entry(rdev1, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
160 161
		int j = rdev1->raid_disk;

162
		if (mddev->level == 10) {
163 164
			/* taking over a raid10-n2 array */
			j /= 2;
165 166
			rdev1->new_raid_disk = j;
		}
167

168 169 170 171 172 173 174 175
		if (mddev->level == 1) {
			/* taiking over a raid1 array-
			 * we have only one active disk
			 */
			j = 0;
			rdev1->new_raid_disk = j;
		}

L
Linus Torvalds 已提交
176
		if (j < 0 || j >= mddev->raid_disks) {
N
NeilBrown 已提交
177 178
			printk(KERN_ERR "md/raid0:%s: bad disk number %d - "
			       "aborting!\n", mdname(mddev), j);
L
Linus Torvalds 已提交
179 180
			goto abort;
		}
181
		if (dev[j]) {
N
NeilBrown 已提交
182 183
			printk(KERN_ERR "md/raid0:%s: multiple devices for %d - "
			       "aborting!\n", mdname(mddev), j);
L
Linus Torvalds 已提交
184 185
			goto abort;
		}
186
		dev[j] = rdev1;
L
Linus Torvalds 已提交
187

188 189
		disk_stack_limits(mddev->gendisk, rdev1->bdev,
				  rdev1->data_offset << 9);
L
Linus Torvalds 已提交
190
		/* as we don't honour merge_bvec_fn, we must never risk
191 192
		 * violating it, so limit ->max_segments to 1, lying within
		 * a single page.
L
Linus Torvalds 已提交
193 194
		 */

195 196 197 198 199
		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn) {
			blk_queue_max_segments(mddev->queue, 1);
			blk_queue_segment_boundary(mddev->queue,
						   PAGE_CACHE_SIZE - 1);
		}
200
		if (!smallest || (rdev1->sectors < smallest->sectors))
L
Linus Torvalds 已提交
201 202 203 204
			smallest = rdev1;
		cnt++;
	}
	if (cnt != mddev->raid_disks) {
N
NeilBrown 已提交
205 206
		printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - "
		       "aborting!\n", mdname(mddev), cnt, mddev->raid_disks);
L
Linus Torvalds 已提交
207 208 209
		goto abort;
	}
	zone->nb_dev = cnt;
210
	zone->zone_end = smallest->sectors * cnt;
L
Linus Torvalds 已提交
211

212
	curr_zone_end = zone->zone_end;
L
Linus Torvalds 已提交
213 214 215 216

	/* now do the other zones */
	for (i = 1; i < conf->nr_strip_zones; i++)
	{
217 218
		int j;

L
Linus Torvalds 已提交
219
		zone = conf->strip_zone + i;
220
		dev = conf->devlist + i * mddev->raid_disks;
L
Linus Torvalds 已提交
221

222
		pr_debug("md/raid0:%s: zone %d\n", mdname(mddev), i);
223
		zone->dev_start = smallest->sectors;
L
Linus Torvalds 已提交
224 225 226 227
		smallest = NULL;
		c = 0;

		for (j=0; j<cnt; j++) {
228
			rdev = conf->devlist[j];
229
			if (rdev->sectors <= zone->dev_start) {
230 231 232
				pr_debug("md/raid0:%s: checking %s ... nope\n",
					 mdname(mddev),
					 bdevname(rdev->bdev, b));
233 234
				continue;
			}
235 236 237 238
			pr_debug("md/raid0:%s: checking %s ..."
				 " contained as device %d\n",
				 mdname(mddev),
				 bdevname(rdev->bdev, b), c);
239
			dev[c] = rdev;
240 241 242
			c++;
			if (!smallest || rdev->sectors < smallest->sectors) {
				smallest = rdev;
243 244 245
				pr_debug("md/raid0:%s:  (%llu) is smallest!.\n",
					 mdname(mddev),
					 (unsigned long long)rdev->sectors);
246
			}
L
Linus Torvalds 已提交
247 248 249
		}

		zone->nb_dev = c;
250
		sectors = (smallest->sectors - zone->dev_start) * c;
251 252 253
		pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n",
			 mdname(mddev),
			 zone->nb_dev, (unsigned long long)sectors);
L
Linus Torvalds 已提交
254

255
		curr_zone_end += sectors;
256
		zone->zone_end = curr_zone_end;
L
Linus Torvalds 已提交
257

258 259 260
		pr_debug("md/raid0:%s: current zone start: %llu\n",
			 mdname(mddev),
			 (unsigned long long)smallest->sectors);
L
Linus Torvalds 已提交
261
	}
262 263
	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
L
Linus Torvalds 已提交
264

265 266 267 268
	/*
	 * now since we have the hard sector sizes, we can make sure
	 * chunk size is a multiple of that sector size
	 */
269
	if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
N
NeilBrown 已提交
270
		printk(KERN_ERR "md/raid0:%s: chunk_size of %d not valid\n",
271
		       mdname(mddev),
272
		       mddev->chunk_sectors << 9);
273 274
		goto abort;
	}
275 276 277 278 279

	blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
	blk_queue_io_opt(mddev->queue,
			 (mddev->chunk_sectors << 9) * mddev->raid_disks);

280
	pr_debug("md/raid0:%s: done.\n", mdname(mddev));
281 282
	*private_conf = conf;

L
Linus Torvalds 已提交
283
	return 0;
284
abort:
285 286 287
	kfree(conf->strip_zone);
	kfree(conf->devlist);
	kfree(conf);
288
	*private_conf = NULL;
289
	return err;
L
Linus Torvalds 已提交
290 291 292 293 294
}

/**
 *	raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
 *	@q: request queue
295
 *	@bvm: properties of new bio
L
Linus Torvalds 已提交
296 297 298 299
 *	@biovec: the request that could be merged to it.
 *
 *	Return amount of bytes we can accept at this offset
 */
300 301 302
static int raid0_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
L
Linus Torvalds 已提交
303 304
{
	mddev_t *mddev = q->queuedata;
305
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
L
Linus Torvalds 已提交
306
	int max;
307
	unsigned int chunk_sectors = mddev->chunk_sectors;
308
	unsigned int bio_sectors = bvm->bi_size >> 9;
L
Linus Torvalds 已提交
309

N
NeilBrown 已提交
310
	if (is_power_of_2(chunk_sectors))
311 312 313 314 315
		max =  (chunk_sectors - ((sector & (chunk_sectors-1))
						+ bio_sectors)) << 9;
	else
		max =  (chunk_sectors - (sector_div(sector, chunk_sectors)
						+ bio_sectors)) << 9;
L
Linus Torvalds 已提交
316 317 318 319 320 321 322
	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	if (max <= biovec->bv_len && bio_sectors == 0)
		return biovec->bv_len;
	else 
		return max;
}

323 324 325 326 327 328 329 330 331 332 333 334 335 336
static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	sector_t array_sectors = 0;
	mdk_rdev_t *rdev;

	WARN_ONCE(sectors || raid_disks,
		  "%s does not support generic reshape\n", __func__);

	list_for_each_entry(rdev, &mddev->disks, same_set)
		array_sectors += rdev->sectors;

	return array_sectors;
}

337
static int raid0_run(mddev_t *mddev)
L
Linus Torvalds 已提交
338
{
339
	raid0_conf_t *conf;
340
	int ret;
L
Linus Torvalds 已提交
341

342
	if (mddev->chunk_sectors == 0) {
N
NeilBrown 已提交
343 344
		printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n",
		       mdname(mddev));
345 346
		return -EINVAL;
	}
347 348
	if (md_check_no_bitmap(mddev))
		return -EINVAL;
349
	blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
L
Linus Torvalds 已提交
350

351 352 353 354 355 356 357 358
	/* if private is not null, we are here after takeover */
	if (mddev->private == NULL) {
		ret = create_strip_zones(mddev, &conf);
		if (ret < 0)
			return ret;
		mddev->private = conf;
	}
	conf = mddev->private;
L
Linus Torvalds 已提交
359 360

	/* calculate array device size */
361
	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
L
Linus Torvalds 已提交
362

N
NeilBrown 已提交
363 364 365
	printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n",
	       mdname(mddev),
	       (unsigned long long)mddev->array_sectors);
L
Linus Torvalds 已提交
366 367 368 369 370 371 372 373 374 375
	/* calculate the max read-ahead size.
	 * For read-ahead of large files to be effective, we need to
	 * readahead at least twice a whole stripe. i.e. number of devices
	 * multiplied by chunk size times 2.
	 * If an individual device has an ra_pages greater than the
	 * chunk size, then we will not drive that device as hard as it
	 * wants.  We consider this a configuration error: a larger
	 * chunksize should be used in that case.
	 */
	{
376 377
		int stripe = mddev->raid_disks *
			(mddev->chunk_sectors << 9) / PAGE_SIZE;
L
Linus Torvalds 已提交
378 379 380 381 382
		if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
			mddev->queue->backing_dev_info.ra_pages = 2* stripe;
	}

	blk_queue_merge_bvec(mddev->queue, raid0_mergeable_bvec);
383
	dump_zones(mddev);
384
	return md_integrity_register(mddev);
L
Linus Torvalds 已提交
385 386
}

387
static int raid0_stop(mddev_t *mddev)
L
Linus Torvalds 已提交
388
{
389
	raid0_conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
390 391

	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
392
	kfree(conf->strip_zone);
393
	kfree(conf->devlist);
394
	kfree(conf);
L
Linus Torvalds 已提交
395 396 397 398
	mddev->private = NULL;
	return 0;
}

399 400 401
/* Find the zone which holds a particular offset
 * Update *sectorp to be an offset in that zone
 */
402
static struct strip_zone *find_zone(struct raid0_private_data *conf,
403
				    sector_t *sectorp)
404 405 406
{
	int i;
	struct strip_zone *z = conf->strip_zone;
407
	sector_t sector = *sectorp;
408 409

	for (i = 0; i < conf->nr_strip_zones; i++)
410 411 412
		if (sector < z[i].zone_end) {
			if (i)
				*sectorp = sector - z[i-1].zone_end;
413
			return z + i;
414
		}
415 416 417
	BUG();
}

418 419 420 421 422 423
/*
 * remaps the bio to the target device. we separate two flows.
 * power 2 flow and a general flow for the sake of perfromance
*/
static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,
				sector_t sector, sector_t *sector_offset)
L
Linus Torvalds 已提交
424
{
425 426
	unsigned int sect_in_chunk;
	sector_t chunk;
427
	raid0_conf_t *conf = mddev->private;
428
	int raid_disks = conf->strip_zone[0].nb_dev;
429
	unsigned int chunk_sects = mddev->chunk_sectors;
430

N
NeilBrown 已提交
431
	if (is_power_of_2(chunk_sects)) {
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
		int chunksect_bits = ffz(~chunk_sects);
		/* find the sector offset inside the chunk */
		sect_in_chunk  = sector & (chunk_sects - 1);
		sector >>= chunksect_bits;
		/* chunk in zone */
		chunk = *sector_offset;
		/* quotient is the chunk in real device*/
		sector_div(chunk, zone->nb_dev << chunksect_bits);
	} else{
		sect_in_chunk = sector_div(sector, chunk_sects);
		chunk = *sector_offset;
		sector_div(chunk, chunk_sects * zone->nb_dev);
	}
	/*
	*  position the bio over the real device
	*  real sector = chunk in device + starting of zone
	*	+ the position in the chunk
	*/
	*sector_offset = (chunk * chunk_sects) + sect_in_chunk;
451
	return conf->devlist[(zone - conf->strip_zone)*raid_disks
452 453 454 455 456 457 458 459 460
			     + sector_div(sector, zone->nb_dev)];
}

/*
 * Is io distribute over 1 or more chunks ?
*/
static inline int is_io_in_chunk_boundary(mddev_t *mddev,
			unsigned int chunk_sects, struct bio *bio)
{
N
NeilBrown 已提交
461
	if (likely(is_power_of_2(chunk_sects))) {
462 463 464 465 466 467 468 469 470
		return chunk_sects >= ((bio->bi_sector & (chunk_sects-1))
					+ (bio->bi_size >> 9));
	} else{
		sector_t sector = bio->bi_sector;
		return chunk_sects >= (sector_div(sector, chunk_sects)
						+ (bio->bi_size >> 9));
	}
}

471
static int raid0_make_request(mddev_t *mddev, struct bio *bio)
472 473 474
{
	unsigned int chunk_sects;
	sector_t sector_offset;
L
Linus Torvalds 已提交
475 476 477
	struct strip_zone *zone;
	mdk_rdev_t *tmp_dev;

T
Tejun Heo 已提交
478 479
	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bio);
480 481 482
		return 0;
	}

483
	chunk_sects = mddev->chunk_sectors;
484 485
	if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) {
		sector_t sector = bio->bi_sector;
L
Linus Torvalds 已提交
486 487 488 489 490 491 492 493
		struct bio_pair *bp;
		/* Sanity check -- queue functions should prevent this happening */
		if (bio->bi_vcnt != 1 ||
		    bio->bi_idx != 0)
			goto bad_map;
		/* This is a one page bio that upper layers
		 * refuse to split for us, so we need to split it.
		 */
N
NeilBrown 已提交
494
		if (likely(is_power_of_2(chunk_sects)))
495 496 497 498 499
			bp = bio_split(bio, chunk_sects - (sector &
							   (chunk_sects-1)));
		else
			bp = bio_split(bio, chunk_sects -
				       sector_div(sector, chunk_sects));
500
		if (raid0_make_request(mddev, &bp->bio1))
L
Linus Torvalds 已提交
501
			generic_make_request(&bp->bio1);
502
		if (raid0_make_request(mddev, &bp->bio2))
L
Linus Torvalds 已提交
503 504 505 506 507 508
			generic_make_request(&bp->bio2);

		bio_pair_release(bp);
		return 0;
	}

509 510 511 512
	sector_offset = bio->bi_sector;
	zone =  find_zone(mddev->private, &sector_offset);
	tmp_dev = map_sector(mddev, zone, bio->bi_sector,
			     &sector_offset);
L
Linus Torvalds 已提交
513
	bio->bi_bdev = tmp_dev->bdev;
514 515
	bio->bi_sector = sector_offset + zone->dev_start +
		tmp_dev->data_offset;
L
Linus Torvalds 已提交
516 517 518 519 520 521
	/*
	 * Let the main block layer submit the IO and resolve recursion:
	 */
	return 1;

bad_map:
N
NeilBrown 已提交
522 523 524 525
	printk("md/raid0:%s: make_request bug: can't convert block across chunks"
	       " or bigger than %dk %llu %d\n",
	       mdname(mddev), chunk_sects / 2,
	       (unsigned long long)bio->bi_sector, bio->bi_size >> 10);
L
Linus Torvalds 已提交
526

527
	bio_io_error(bio);
L
Linus Torvalds 已提交
528 529
	return 0;
}
N
NeilBrown 已提交
530

531
static void raid0_status(struct seq_file *seq, mddev_t *mddev)
L
Linus Torvalds 已提交
532
{
533
	seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
L
Linus Torvalds 已提交
534 535 536
	return;
}

M
Maciej Trela 已提交
537
static void *raid0_takeover_raid45(mddev_t *mddev)
538 539 540 541 542
{
	mdk_rdev_t *rdev;
	raid0_conf_t *priv_conf;

	if (mddev->degraded != 1) {
N
NeilBrown 已提交
543 544
		printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
		       mdname(mddev),
545 546 547 548 549 550 551
		       mddev->degraded);
		return ERR_PTR(-EINVAL);
	}

	list_for_each_entry(rdev, &mddev->disks, same_set) {
		/* check slot number for a disk */
		if (rdev->raid_disk == mddev->raid_disks-1) {
N
NeilBrown 已提交
552 553
			printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n",
			       mdname(mddev));
554 555 556 557 558 559
			return ERR_PTR(-EINVAL);
		}
	}

	/* Set new parameters */
	mddev->new_level = 0;
560
	mddev->new_layout = 0;
561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
	mddev->new_chunk_sectors = mddev->chunk_sectors;
	mddev->raid_disks--;
	mddev->delta_disks = -1;
	/* make sure it will be not marked as dirty */
	mddev->recovery_cp = MaxSector;

	create_strip_zones(mddev, &priv_conf);
	return priv_conf;
}

static void *raid0_takeover_raid10(mddev_t *mddev)
{
	raid0_conf_t *priv_conf;

	/* Check layout:
	 *  - far_copies must be 1
	 *  - near_copies must be 2
	 *  - disks number must be even
	 *  - all mirrors must be already degraded
	 */
	if (mddev->layout != ((1 << 8) + 2)) {
N
NeilBrown 已提交
582 583
		printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takover layout: 0x%x\n",
		       mdname(mddev),
584 585 586 587
		       mddev->layout);
		return ERR_PTR(-EINVAL);
	}
	if (mddev->raid_disks & 1) {
N
NeilBrown 已提交
588 589
		printk(KERN_ERR "md/raid0:%s: Raid0 cannot takover Raid10 with odd disk number.\n",
		       mdname(mddev));
590 591 592
		return ERR_PTR(-EINVAL);
	}
	if (mddev->degraded != (mddev->raid_disks>>1)) {
N
NeilBrown 已提交
593 594
		printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n",
		       mdname(mddev));
595 596 597 598 599
		return ERR_PTR(-EINVAL);
	}

	/* Set new parameters */
	mddev->new_level = 0;
600
	mddev->new_layout = 0;
601 602 603 604 605 606 607 608 609 610 611
	mddev->new_chunk_sectors = mddev->chunk_sectors;
	mddev->delta_disks = - mddev->raid_disks / 2;
	mddev->raid_disks += mddev->delta_disks;
	mddev->degraded = 0;
	/* make sure it will be not marked as dirty */
	mddev->recovery_cp = MaxSector;

	create_strip_zones(mddev, &priv_conf);
	return priv_conf;
}

612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629
static void *raid0_takeover_raid1(mddev_t *mddev)
{
	raid0_conf_t *priv_conf;

	/* Check layout:
	 *  - (N - 1) mirror drives must be already faulty
	 */
	if ((mddev->raid_disks - 1) != mddev->degraded) {
		printk(KERN_ERR "md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
		       mdname(mddev));
		return ERR_PTR(-EINVAL);
	}

	/* Set new parameters */
	mddev->new_level = 0;
	mddev->new_layout = 0;
	mddev->new_chunk_sectors = 128; /* by default set chunk size to 64k */
	mddev->delta_disks = 1 - mddev->raid_disks;
K
Krzysztof Wojcik 已提交
630
	mddev->raid_disks = 1;
631 632 633 634 635 636 637
	/* make sure it will be not marked as dirty */
	mddev->recovery_cp = MaxSector;

	create_strip_zones(mddev, &priv_conf);
	return priv_conf;
}

638 639 640
static void *raid0_takeover(mddev_t *mddev)
{
	/* raid0 can take over:
M
Maciej Trela 已提交
641
	 *  raid4 - if all data disks are active.
642 643
	 *  raid5 - providing it is Raid4 layout and one disk is faulty
	 *  raid10 - assuming we have all necessary active disks
644
	 *  raid1 - with (N -1) mirror drives faulty
645
	 */
M
Maciej Trela 已提交
646 647 648
	if (mddev->level == 4)
		return raid0_takeover_raid45(mddev);

649 650
	if (mddev->level == 5) {
		if (mddev->layout == ALGORITHM_PARITY_N)
M
Maciej Trela 已提交
651
			return raid0_takeover_raid45(mddev);
652

N
NeilBrown 已提交
653 654
		printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
		       mdname(mddev), ALGORITHM_PARITY_N);
655 656 657 658 659
	}

	if (mddev->level == 10)
		return raid0_takeover_raid10(mddev);

660 661 662 663 664 665
	if (mddev->level == 1)
		return raid0_takeover_raid1(mddev);

	printk(KERN_ERR "Takeover from raid%i to raid0 not supported\n",
		mddev->level);

666 667 668 669 670 671 672
	return ERR_PTR(-EINVAL);
}

static void raid0_quiesce(mddev_t *mddev, int state)
{
}

673
static struct mdk_personality raid0_personality=
L
Linus Torvalds 已提交
674 675
{
	.name		= "raid0",
676
	.level		= 0,
L
Linus Torvalds 已提交
677 678 679 680 681
	.owner		= THIS_MODULE,
	.make_request	= raid0_make_request,
	.run		= raid0_run,
	.stop		= raid0_stop,
	.status		= raid0_status,
682
	.size		= raid0_size,
683 684
	.takeover	= raid0_takeover,
	.quiesce	= raid0_quiesce,
L
Linus Torvalds 已提交
685 686 687 688
};

static int __init raid0_init (void)
{
689
	return register_md_personality (&raid0_personality);
L
Linus Torvalds 已提交
690 691 692 693
}

static void raid0_exit (void)
{
694
	unregister_md_personality (&raid0_personality);
L
Linus Torvalds 已提交
695 696 697 698 699
}

module_init(raid0_init);
module_exit(raid0_exit);
MODULE_LICENSE("GPL");
700
MODULE_DESCRIPTION("RAID0 (striping) personality for MD");
L
Linus Torvalds 已提交
701
MODULE_ALIAS("md-personality-2"); /* RAID0 */
702
MODULE_ALIAS("md-raid0");
703
MODULE_ALIAS("md-level-0");