raid0.c 14.1 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
/*
   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.  
*/

#include <linux/module.h>
#include <linux/raid/raid0.h>

#define MAJOR_NR MD_MAJOR
#define MD_DRIVER
#define MD_PERSONALITY

28
static void raid0_unplug(struct request_queue *q)
Linus Torvalds's avatar
Linus Torvalds committed
29 30 31 32 33 34 35
{
	mddev_t *mddev = q->queuedata;
	raid0_conf_t *conf = mddev_to_conf(mddev);
	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	int i;

	for (i=0; i<mddev->raid_disks; i++) {
36
		struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);
Linus Torvalds's avatar
Linus Torvalds committed
37

38
		blk_unplug(r_queue);
Linus Torvalds's avatar
Linus Torvalds committed
39 40 41
	}
}

42 43 44 45 46 47 48 49
static int raid0_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	raid0_conf_t *conf = mddev_to_conf(mddev);
	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
	int i, ret = 0;

	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
50
		struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
51 52 53 54 55 56

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

Linus Torvalds's avatar
Linus Torvalds committed
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74

static int create_strip_zones (mddev_t *mddev)
{
	int i, c, j;
	sector_t current_offset, curr_zone_offset;
	sector_t min_spacing;
	raid0_conf_t *conf = mddev_to_conf(mddev);
	mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev;
	struct list_head *tmp1, *tmp2;
	struct strip_zone *zone;
	int cnt;
	char b[BDEVNAME_SIZE];
 
	/*
	 * The number of 'same size groups'
	 */
	conf->nr_strip_zones = 0;
 
75
	rdev_for_each(rdev1, tmp1, mddev) {
Linus Torvalds's avatar
Linus Torvalds committed
76 77 78
		printk("raid0: looking at %s\n",
			bdevname(rdev1->bdev,b));
		c = 0;
79
		rdev_for_each(rdev2, tmp2, mddev) {
Linus Torvalds's avatar
Linus Torvalds committed
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109
			printk("raid0:   comparing %s(%llu)",
			       bdevname(rdev1->bdev,b),
			       (unsigned long long)rdev1->size);
			printk(" with %s(%llu)\n",
			       bdevname(rdev2->bdev,b),
			       (unsigned long long)rdev2->size);
			if (rdev2 == rdev1) {
				printk("raid0:   END\n");
				break;
			}
			if (rdev2->size == rdev1->size)
			{
				/*
				 * Not unique, don't count it as a new
				 * group
				 */
				printk("raid0:   EQUAL\n");
				c = 1;
				break;
			}
			printk("raid0:   NOT EQUAL\n");
		}
		if (!c) {
			printk("raid0:   ==> UNIQUE\n");
			conf->nr_strip_zones++;
			printk("raid0: %d zones\n", conf->nr_strip_zones);
		}
	}
	printk("raid0: FINAL %d zones\n", conf->nr_strip_zones);

110
	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
Linus Torvalds's avatar
Linus Torvalds committed
111 112 113
				conf->nr_strip_zones, GFP_KERNEL);
	if (!conf->strip_zone)
		return 1;
114
	conf->devlist = kzalloc(sizeof(mdk_rdev_t*)*
Linus Torvalds's avatar
Linus Torvalds committed
115 116 117 118 119 120 121 122 123 124 125 126
				conf->nr_strip_zones*mddev->raid_disks,
				GFP_KERNEL);
	if (!conf->devlist)
		return 1;

	/* 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;
	zone->dev = conf->devlist;
127
	rdev_for_each(rdev1, tmp1, mddev) {
Linus Torvalds's avatar
Linus Torvalds committed
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
		int j = rdev1->raid_disk;

		if (j < 0 || j >= mddev->raid_disks) {
			printk("raid0: bad disk number %d - aborting!\n", j);
			goto abort;
		}
		if (zone->dev[j]) {
			printk("raid0: multiple devices for %d - aborting!\n",
				j);
			goto abort;
		}
		zone->dev[j] = rdev1;

		blk_queue_stack_limits(mddev->queue,
				       rdev1->bdev->bd_disk->queue);
		/* as we don't honour merge_bvec_fn, we must never risk
		 * violating it, so limit ->max_sector to one PAGE, as
		 * a one page request is never in violation.
		 */

		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn &&
		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

		if (!smallest || (rdev1->size <smallest->size))
			smallest = rdev1;
		cnt++;
	}
	if (cnt != mddev->raid_disks) {
		printk("raid0: too few disks (%d of %d) - aborting!\n",
			cnt, mddev->raid_disks);
		goto abort;
	}
	zone->nb_dev = cnt;
	zone->size = smallest->size * cnt;
	zone->zone_offset = 0;

	current_offset = smallest->size;
	curr_zone_offset = zone->size;

	/* now do the other zones */
	for (i = 1; i < conf->nr_strip_zones; i++)
	{
		zone = conf->strip_zone + i;
		zone->dev = conf->strip_zone[i-1].dev + mddev->raid_disks;

		printk("raid0: zone %d\n", i);
		zone->dev_offset = current_offset;
		smallest = NULL;
		c = 0;

		for (j=0; j<cnt; j++) {
			char b[BDEVNAME_SIZE];
			rdev = conf->strip_zone[0].dev[j];
			printk("raid0: checking %s ...", bdevname(rdev->bdev,b));
			if (rdev->size > current_offset)
			{
				printk(" contained as device %d\n", c);
				zone->dev[c] = rdev;
				c++;
				if (!smallest || (rdev->size <smallest->size)) {
					smallest = rdev;
					printk("  (%llu) is smallest!.\n", 
						(unsigned long long)rdev->size);
				}
			} else
				printk(" nope.\n");
		}

		zone->nb_dev = c;
		zone->size = (smallest->size - current_offset) * c;
		printk("raid0: zone->nb_dev: %d, size: %llu\n",
			zone->nb_dev, (unsigned long long)zone->size);

		zone->zone_offset = curr_zone_offset;
		curr_zone_offset += zone->size;

		current_offset = smallest->size;
		printk("raid0: current zone offset: %llu\n",
			(unsigned long long)current_offset);
	}

	/* Now find appropriate hash spacing.
	 * We want a number which causes most hash entries to cover
	 * at most two strips, but the hash table must be at most
	 * 1 PAGE.  We choose the smallest strip, or contiguous collection
	 * of strips, that has big enough size.  We never consider the last
	 * strip though as it's size has no bearing on the efficacy of the hash
	 * table.
	 */
	conf->hash_spacing = curr_zone_offset;
	min_spacing = curr_zone_offset;
	sector_div(min_spacing, PAGE_SIZE/sizeof(struct strip_zone*));
	for (i=0; i < conf->nr_strip_zones-1; i++) {
		sector_t sz = 0;
		for (j=i; j<conf->nr_strip_zones-1 &&
			     sz < min_spacing ; j++)
			sz += conf->strip_zone[j].size;
		if (sz >= min_spacing && sz < conf->hash_spacing)
			conf->hash_spacing = sz;
	}

	mddev->queue->unplug_fn = raid0_unplug;

232 233
	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
Linus Torvalds's avatar
Linus Torvalds committed
234 235 236 237 238 239 240 241 242 243

	printk("raid0: done.\n");
	return 0;
 abort:
	return 1;
}

/**
 *	raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
 *	@q: request queue
244
 *	@bvm: properties of new bio
Linus Torvalds's avatar
Linus Torvalds committed
245 246 247 248
 *	@biovec: the request that could be merged to it.
 *
 *	Return amount of bytes we can accept at this offset
 */
249 250 251
static int raid0_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
Linus Torvalds's avatar
Linus Torvalds committed
252 253
{
	mddev_t *mddev = q->queuedata;
254
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
Linus Torvalds's avatar
Linus Torvalds committed
255 256
	int max;
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
257
	unsigned int bio_sectors = bvm->bi_size >> 9;
Linus Torvalds's avatar
Linus Torvalds committed
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274

	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	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;
}

static int raid0_run (mddev_t *mddev)
{
	unsigned  cur=0, i=0, nb_zone;
	s64 size;
	raid0_conf_t *conf;
	mdk_rdev_t *rdev;
	struct list_head *tmp;

275 276 277 278 279
	if (mddev->chunk_size == 0) {
		printk(KERN_ERR "md/raid0: non-zero chunk size required.\n");
		return -EINVAL;
	}
	printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n",
Linus Torvalds's avatar
Linus Torvalds committed
280 281 282 283 284
	       mdname(mddev),
	       mddev->chunk_size >> 9,
	       (mddev->chunk_size>>1)-1);
	blk_queue_max_sectors(mddev->queue, mddev->chunk_size >> 9);
	blk_queue_segment_boundary(mddev->queue, (mddev->chunk_size>>1) - 1);
285
	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
Linus Torvalds's avatar
Linus Torvalds committed
286 287 288 289 290 291 292 293 294 295 296 297

	conf = kmalloc(sizeof (raid0_conf_t), GFP_KERNEL);
	if (!conf)
		goto out;
	mddev->private = (void *)conf;
 
	conf->strip_zone = NULL;
	conf->devlist = NULL;
	if (create_strip_zones (mddev)) 
		goto out_free_conf;

	/* calculate array device size */
298
	mddev->array_sectors = 0;
299
	rdev_for_each(rdev, tmp, mddev)
300
		mddev->array_sectors += rdev->size * 2;
Linus Torvalds's avatar
Linus Torvalds committed
301 302

	printk("raid0 : md_size is %llu blocks.\n", 
303
		(unsigned long long)mddev->array_sectors / 2);
Linus Torvalds's avatar
Linus Torvalds committed
304 305 306
	printk("raid0 : conf->hash_spacing is %llu blocks.\n",
		(unsigned long long)conf->hash_spacing);
	{
307
		sector_t s = mddev->array_sectors / 2;
Linus Torvalds's avatar
Linus Torvalds committed
308 309 310
		sector_t space = conf->hash_spacing;
		int round;
		conf->preshift = 0;
311
		if (sizeof(sector_t) > sizeof(u32)) {
Linus Torvalds's avatar
Linus Torvalds committed
312
			/*shift down space and s so that sector_div will work */
313
			while (space > (sector_t) (~(u32)0)) {
Linus Torvalds's avatar
Linus Torvalds committed
314 315 316 317 318 319
				s >>= 1;
				space >>= 1;
				s += 1; /* force round-up */
				conf->preshift++;
			}
		}
320
		round = sector_div(s, (u32)space) ? 1 : 0;
Linus Torvalds's avatar
Linus Torvalds committed
321 322 323 324 325 326 327 328 329 330 331
		nb_zone = s + round;
	}
	printk("raid0 : nb_zone is %d.\n", nb_zone);

	printk("raid0 : Allocating %Zd bytes for hash.\n",
				nb_zone*sizeof(struct strip_zone*));
	conf->hash_table = kmalloc (sizeof (struct strip_zone *)*nb_zone, GFP_KERNEL);
	if (!conf->hash_table)
		goto out_free_conf;
	size = conf->strip_zone[cur].size;

332 333
	conf->hash_table[0] = conf->strip_zone + cur;
	for (i=1; i< nb_zone; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
334 335 336 337 338
		while (size <= conf->hash_spacing) {
			cur++;
			size += conf->strip_zone[cur].size;
		}
		size -= conf->hash_spacing;
339
		conf->hash_table[i] = conf->strip_zone + cur;
Linus Torvalds's avatar
Linus Torvalds committed
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358
	}
	if (conf->preshift) {
		conf->hash_spacing >>= conf->preshift;
		/* round hash_spacing up so when we divide by it, we
		 * err on the side of too-low, which is safest
		 */
		conf->hash_spacing++;
	}

	/* 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.
	 */
	{
359
		int stripe = mddev->raid_disks * mddev->chunk_size / PAGE_SIZE;
Linus Torvalds's avatar
Linus Torvalds committed
360 361 362 363 364 365 366 367 368
		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);
	return 0;

out_free_conf:
369 370
	kfree(conf->strip_zone);
	kfree(conf->devlist);
Linus Torvalds's avatar
Linus Torvalds committed
371 372 373
	kfree(conf);
	mddev->private = NULL;
out:
374
	return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
375 376 377 378 379 380 381
}

static int raid0_stop (mddev_t *mddev)
{
	raid0_conf_t *conf = mddev_to_conf(mddev);

	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
382
	kfree(conf->hash_table);
Linus Torvalds's avatar
Linus Torvalds committed
383
	conf->hash_table = NULL;
384
	kfree(conf->strip_zone);
Linus Torvalds's avatar
Linus Torvalds committed
385
	conf->strip_zone = NULL;
386
	kfree(conf);
Linus Torvalds's avatar
Linus Torvalds committed
387 388 389 390 391
	mddev->private = NULL;

	return 0;
}

392
static int raid0_make_request (struct request_queue *q, struct bio *bio)
Linus Torvalds's avatar
Linus Torvalds committed
393 394 395 396 397 398
{
	mddev_t *mddev = q->queuedata;
	unsigned int sect_in_chunk, chunksize_bits,  chunk_size, chunk_sects;
	raid0_conf_t *conf = mddev_to_conf(mddev);
	struct strip_zone *zone;
	mdk_rdev_t *tmp_dev;
399
	sector_t chunk;
Linus Torvalds's avatar
Linus Torvalds committed
400
	sector_t block, rsect;
401
	const int rw = bio_data_dir(bio);
Tejun Heo's avatar
Tejun Heo committed
402
	int cpu;
Linus Torvalds's avatar
Linus Torvalds committed
403

404
	if (unlikely(bio_barrier(bio))) {
405
		bio_endio(bio, -EOPNOTSUPP);
406 407 408
		return 0;
	}

409 410 411 412 413
	cpu = part_stat_lock();
	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
		      bio_sectors(bio));
	part_stat_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442

	chunk_size = mddev->chunk_size >> 10;
	chunk_sects = mddev->chunk_size >> 9;
	chunksize_bits = ffz(~chunk_size);
	block = bio->bi_sector >> 1;
	

	if (unlikely(chunk_sects < (bio->bi_sector & (chunk_sects - 1)) + (bio->bi_size >> 9))) {
		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.
		 */
		bp = bio_split(bio, bio_split_pool, chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
		if (raid0_make_request(q, &bp->bio1))
			generic_make_request(&bp->bio1);
		if (raid0_make_request(q, &bp->bio2))
			generic_make_request(&bp->bio2);

		bio_pair_release(bp);
		return 0;
	}
 

	{
		sector_t x = block >> conf->preshift;
443
		sector_div(x, (u32)conf->hash_spacing);
Linus Torvalds's avatar
Linus Torvalds committed
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477
		zone = conf->hash_table[x];
	}
 
	while (block >= (zone->zone_offset + zone->size)) 
		zone++;
    
	sect_in_chunk = bio->bi_sector & ((chunk_size<<1) -1);


	{
		sector_t x =  (block - zone->zone_offset) >> chunksize_bits;

		sector_div(x, zone->nb_dev);
		chunk = x;

		x = block >> chunksize_bits;
		tmp_dev = zone->dev[sector_div(x, zone->nb_dev)];
	}
	rsect = (((chunk << chunksize_bits) + zone->dev_offset)<<1)
		+ sect_in_chunk;
 
	bio->bi_bdev = tmp_dev->bdev;
	bio->bi_sector = rsect + tmp_dev->data_offset;

	/*
	 * Let the main block layer submit the IO and resolve recursion:
	 */
	return 1;

bad_map:
	printk("raid0_make_request bug: can't convert block across chunks"
		" or bigger than %dk %llu %d\n", chunk_size, 
		(unsigned long long)bio->bi_sector, bio->bi_size >> 10);

478
	bio_io_error(bio);
Linus Torvalds's avatar
Linus Torvalds committed
479 480
	return 0;
}
481

Linus Torvalds's avatar
Linus Torvalds committed
482 483 484 485 486 487 488
static void raid0_status (struct seq_file *seq, mddev_t *mddev)
{
#undef MD_DEBUG
#ifdef MD_DEBUG
	int j, k, h;
	char b[BDEVNAME_SIZE];
	raid0_conf_t *conf = mddev_to_conf(mddev);
489

Linus Torvalds's avatar
Linus Torvalds committed
490 491 492 493
	h = 0;
	for (j = 0; j < conf->nr_strip_zones; j++) {
		seq_printf(seq, "      z%d", j);
		if (conf->hash_table[h] == conf->strip_zone+j)
494
			seq_printf(seq, "(h%d)", h++);
Linus Torvalds's avatar
Linus Torvalds committed
495 496
		seq_printf(seq, "=[");
		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
497
			seq_printf(seq, "%s/", bdevname(
Linus Torvalds's avatar
Linus Torvalds committed
498 499
				conf->strip_zone[j].dev[k]->bdev,b));

500
		seq_printf(seq, "] zo=%d do=%d s=%d\n",
Linus Torvalds's avatar
Linus Torvalds committed
501 502 503 504 505 506 507 508 509
				conf->strip_zone[j].zone_offset,
				conf->strip_zone[j].dev_offset,
				conf->strip_zone[j].size);
	}
#endif
	seq_printf(seq, " %dk chunks", mddev->chunk_size/1024);
	return;
}

510
static struct mdk_personality raid0_personality=
Linus Torvalds's avatar
Linus Torvalds committed
511 512
{
	.name		= "raid0",
513
	.level		= 0,
Linus Torvalds's avatar
Linus Torvalds committed
514 515 516 517 518 519 520 521 522
	.owner		= THIS_MODULE,
	.make_request	= raid0_make_request,
	.run		= raid0_run,
	.stop		= raid0_stop,
	.status		= raid0_status,
};

static int __init raid0_init (void)
{
523
	return register_md_personality (&raid0_personality);
Linus Torvalds's avatar
Linus Torvalds committed
524 525 526 527
}

static void raid0_exit (void)
{
528
	unregister_md_personality (&raid0_personality);
Linus Torvalds's avatar
Linus Torvalds committed
529 530 531 532 533 534
}

module_init(raid0_init);
module_exit(raid0_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-2"); /* RAID0 */
535
MODULE_ALIAS("md-raid0");
536
MODULE_ALIAS("md-level-0");