linear.c 10.5 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
/*
   linear.c : Multiple Devices driver for Linux
	      Copyright (C) 1994-96 Marc ZYNGIER
	      <zyngier@ufr-info-p7.ibp.fr> or
	      <maz@gloups.fdn.fr>

   Linear mode 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.  
*/

19 20 21
#include <linux/blkdev.h>
#include <linux/raid/md_u.h>
#include <linux/seq_file.h>
22
#include "md.h"
23
#include "linear.h"
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31

/*
 * find which device holds a particular offset 
 */
static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector)
{
	dev_info_t *hash;
	linear_conf_t *conf = mddev_to_conf(mddev);
32
	sector_t idx = sector >> conf->sector_shift;
L
Linus Torvalds 已提交
33 34 35 36

	/*
	 * sector_div(a,b) returns the remainer and sets a to a/b
	 */
37 38
	(void)sector_div(idx, conf->spacing);
	hash = conf->hash_table[idx];
L
Linus Torvalds 已提交
39

40
	while (sector >= hash->num_sectors + hash->start_sector)
L
Linus Torvalds 已提交
41 42 43 44 45
		hash++;
	return hash;
}

/**
46
 *	linear_mergeable_bvec -- tell bio layer if two requests can be merged
L
Linus Torvalds 已提交
47
 *	@q: request queue
48
 *	@bvm: properties of new bio
L
Linus Torvalds 已提交
49 50 51 52
 *	@biovec: the request that could be merged to it.
 *
 *	Return amount of bytes we can take at this offset
 */
53 54 55
static int linear_mergeable_bvec(struct request_queue *q,
				 struct bvec_merge_data *bvm,
				 struct bio_vec *biovec)
L
Linus Torvalds 已提交
56 57 58
{
	mddev_t *mddev = q->queuedata;
	dev_info_t *dev0;
59 60
	unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9;
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
L
Linus Torvalds 已提交
61 62

	dev0 = which_dev(mddev, sector);
63
	maxsectors = dev0->num_sectors - (sector - dev0->start_sector);
L
Linus Torvalds 已提交
64 65 66 67 68 69 70 71 72 73 74 75 76 77 78

	if (maxsectors < bio_sectors)
		maxsectors = 0;
	else
		maxsectors -= bio_sectors;

	if (maxsectors <= (PAGE_SIZE >> 9 ) && bio_sectors == 0)
		return biovec->bv_len;
	/* The bytes available at this offset could be really big,
	 * so we cap at 2^31 to avoid overflow */
	if (maxsectors > (1 << (31-9)))
		return 1<<31;
	return maxsectors << 9;
}

79
static void linear_unplug(struct request_queue *q)
L
Linus Torvalds 已提交
80 81 82 83 84 85
{
	mddev_t *mddev = q->queuedata;
	linear_conf_t *conf = mddev_to_conf(mddev);
	int i;

	for (i=0; i < mddev->raid_disks; i++) {
86
		struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev);
87
		blk_unplug(r_queue);
L
Linus Torvalds 已提交
88 89 90
	}
}

91 92 93 94 95 96 97
static int linear_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	linear_conf_t *conf = mddev_to_conf(mddev);
	int i, ret = 0;

	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
98
		struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev);
99 100 101 102 103
		ret |= bdi_congested(&q->backing_dev_info, bits);
	}
	return ret;
}

104 105 106 107 108 109 110 111 112 113
static sector_t linear_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	linear_conf_t *conf = mddev_to_conf(mddev);

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

	return conf->array_sectors;
}

114
static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
L
Linus Torvalds 已提交
115 116 117 118 119
{
	linear_conf_t *conf;
	dev_info_t **table;
	mdk_rdev_t *rdev;
	int i, nb_zone, cnt;
120
	sector_t min_sectors;
121
	sector_t curr_sector;
L
Linus Torvalds 已提交
122

123
	conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t),
L
Linus Torvalds 已提交
124 125
			GFP_KERNEL);
	if (!conf)
126 127
		return NULL;

L
Linus Torvalds 已提交
128
	cnt = 0;
129
	conf->array_sectors = 0;
L
Linus Torvalds 已提交
130

131
	list_for_each_entry(rdev, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
132 133 134
		int j = rdev->raid_disk;
		dev_info_t *disk = conf->disks + j;

135
		if (j < 0 || j >= raid_disks || disk->rdev) {
L
Linus Torvalds 已提交
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
			printk("linear: disk numbering problem. Aborting!\n");
			goto out;
		}

		disk->rdev = rdev;

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

152 153
		disk->num_sectors = rdev->sectors;
		conf->array_sectors += rdev->sectors;
L
Linus Torvalds 已提交
154 155 156

		cnt++;
	}
157
	if (cnt != raid_disks) {
L
Linus Torvalds 已提交
158 159 160 161
		printk("linear: not enough drives present. Aborting!\n");
		goto out;
	}

162 163
	min_sectors = conf->array_sectors;
	sector_div(min_sectors, PAGE_SIZE/sizeof(struct dev_info *));
164 165
	if (min_sectors == 0)
		min_sectors = 1;
166

167
	/* min_sectors is the minimum spacing that will fit the hash
168 169
	 * table in one PAGE.  This may be much smaller than needed.
	 * We find the smallest non-terminal set of consecutive devices
170
	 * that is larger than min_sectors and use the size of that as
171 172
	 * the actual spacing
	 */
173
	conf->spacing = conf->array_sectors;
174
	for (i=0; i < cnt-1 ; i++) {
175
		sector_t tmp = 0;
176
		int j;
177 178
		for (j = i; j < cnt - 1 && tmp < min_sectors; j++)
			tmp += conf->disks[j].num_sectors;
179 180
		if (tmp >= min_sectors && tmp < conf->spacing)
			conf->spacing = tmp;
181 182
	}

183
	/* spacing may be too large for sector_div to work with,
184 185
	 * so we might need to pre-shift
	 */
186
	conf->sector_shift = 0;
187
	if (sizeof(sector_t) > sizeof(u32)) {
188
		sector_t space = conf->spacing;
189 190
		while (space > (sector_t)(~(u32)0)) {
			space >>= 1;
191
			conf->sector_shift++;
192 193
		}
	}
L
Linus Torvalds 已提交
194 195 196 197 198 199 200 201 202
	/*
	 * This code was restructured to work around a gcc-2.95.3 internal
	 * compiler error.  Alter it with care.
	 */
	{
		sector_t sz;
		unsigned round;
		unsigned long base;

203
		sz = conf->array_sectors >> conf->sector_shift;
204
		sz += 1; /* force round-up */
205
		base = conf->spacing >> conf->sector_shift;
L
Linus Torvalds 已提交
206
		round = sector_div(sz, base);
207
		nb_zone = sz + (round ? 1 : 0);
L
Linus Torvalds 已提交
208
	}
209 210 211
	BUG_ON(nb_zone > PAGE_SIZE / sizeof(struct dev_info *));

	conf->hash_table = kmalloc (sizeof (struct dev_info *) * nb_zone,
L
Linus Torvalds 已提交
212 213 214 215 216 217
					GFP_KERNEL);
	if (!conf->hash_table)
		goto out;

	/*
	 * Here we generate the linear hash table
218
	 * First calculate the device offsets.
L
Linus Torvalds 已提交
219
	 */
220
	conf->disks[0].start_sector = 0;
221
	for (i = 1; i < raid_disks; i++)
222 223 224
		conf->disks[i].start_sector =
			conf->disks[i-1].start_sector +
			conf->disks[i-1].num_sectors;
225

L
Linus Torvalds 已提交
226
	table = conf->hash_table;
227
	i = 0;
228 229
	for (curr_sector = 0;
	     curr_sector < conf->array_sectors;
230
	     curr_sector += conf->spacing) {
L
Linus Torvalds 已提交
231

232
		while (i < raid_disks-1 &&
233
		       curr_sector >= conf->disks[i+1].start_sector)
234
			i++;
L
Linus Torvalds 已提交
235

236 237 238
		*table ++ = conf->disks + i;
	}

239 240 241
	if (conf->sector_shift) {
		conf->spacing >>= conf->sector_shift;
		/* round spacing up so that when we divide by it,
242
		 * we err on the side of "too-low", which is safest.
L
Linus Torvalds 已提交
243
		 */
244
		conf->spacing++;
L
Linus Torvalds 已提交
245
	}
246 247

	BUG_ON(table - conf->hash_table > nb_zone);
L
Linus Torvalds 已提交
248

249 250 251 252 253 254 255 256 257 258 259
	return conf;

out:
	kfree(conf);
	return NULL;
}

static int linear_run (mddev_t *mddev)
{
	linear_conf_t *conf;

260
	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
261 262 263 264 265
	conf = linear_conf(mddev, mddev->raid_disks);

	if (!conf)
		return 1;
	mddev->private = conf;
266
	mddev->array_sectors = linear_size(mddev, 0, 0);
267

L
Linus Torvalds 已提交
268 269
	blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec);
	mddev->queue->unplug_fn = linear_unplug;
270 271
	mddev->queue->backing_dev_info.congested_fn = linear_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
L
Linus Torvalds 已提交
272
	return 0;
273
}
L
Linus Torvalds 已提交
274

275 276 277 278 279 280 281 282 283 284 285 286
static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
{
	/* Adding a drive to a linear array allows the array to grow.
	 * It is permitted if the new drive has a matching superblock
	 * already on it, with raid_disk equal to raid_disks.
	 * It is achieved by creating a new linear_private_data structure
	 * and swapping it in in-place of the current one.
	 * The current one is never freed until the array is stopped.
	 * This avoids races.
	 */
	linear_conf_t *newconf;

287
	if (rdev->saved_raid_disk != mddev->raid_disks)
288 289
		return -EINVAL;

290 291
	rdev->raid_disk = rdev->saved_raid_disk;

292 293 294 295 296 297 298 299
	newconf = linear_conf(mddev,mddev->raid_disks+1);

	if (!newconf)
		return -ENOMEM;

	newconf->prev = mddev_to_conf(mddev);
	mddev->private = newconf;
	mddev->raid_disks++;
300
	mddev->array_sectors = linear_size(mddev, 0, 0);
301
	set_capacity(mddev->gendisk, mddev->array_sectors);
302
	return 0;
L
Linus Torvalds 已提交
303 304 305 306 307 308 309
}

static int linear_stop (mddev_t *mddev)
{
	linear_conf_t *conf = mddev_to_conf(mddev);
  
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
310 311 312 313 314 315
	do {
		linear_conf_t *t = conf->prev;
		kfree(conf->hash_table);
		kfree(conf);
		conf = t;
	} while (conf);
L
Linus Torvalds 已提交
316 317 318 319

	return 0;
}

320
static int linear_make_request (struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
321
{
322
	const int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
323 324
	mddev_t *mddev = q->queuedata;
	dev_info_t *tmp_dev;
T
Tejun Heo 已提交
325
	int cpu;
L
Linus Torvalds 已提交
326

327
	if (unlikely(bio_barrier(bio))) {
328
		bio_endio(bio, -EOPNOTSUPP);
329 330 331
		return 0;
	}

T
Tejun Heo 已提交
332 333 334 335 336
	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();
L
Linus Torvalds 已提交
337 338 339

	tmp_dev = which_dev(mddev, bio->bi_sector);
    
340 341 342 343
	if (unlikely(bio->bi_sector >= (tmp_dev->num_sectors +
					tmp_dev->start_sector)
		     || (bio->bi_sector <
			 tmp_dev->start_sector))) {
L
Linus Torvalds 已提交
344 345
		char b[BDEVNAME_SIZE];

346 347 348
		printk("linear_make_request: Sector %llu out of bounds on "
			"dev %s: %llu sectors, offset %llu\n",
			(unsigned long long)bio->bi_sector,
L
Linus Torvalds 已提交
349
			bdevname(tmp_dev->rdev->bdev, b),
350 351
			(unsigned long long)tmp_dev->num_sectors,
			(unsigned long long)tmp_dev->start_sector);
352
		bio_io_error(bio);
L
Linus Torvalds 已提交
353 354 355
		return 0;
	}
	if (unlikely(bio->bi_sector + (bio->bi_size >> 9) >
356
		     tmp_dev->start_sector + tmp_dev->num_sectors)) {
L
Linus Torvalds 已提交
357 358 359 360
		/* This bio crosses a device boundary, so we have to
		 * split it.
		 */
		struct bio_pair *bp;
361

D
Denis ChengRq 已提交
362
		bp = bio_split(bio,
363 364 365
			       tmp_dev->start_sector + tmp_dev->num_sectors
			       - bio->bi_sector);

L
Linus Torvalds 已提交
366 367 368 369 370 371 372 373 374
		if (linear_make_request(q, &bp->bio1))
			generic_make_request(&bp->bio1);
		if (linear_make_request(q, &bp->bio2))
			generic_make_request(&bp->bio2);
		bio_pair_release(bp);
		return 0;
	}
		    
	bio->bi_bdev = tmp_dev->rdev->bdev;
375 376
	bio->bi_sector = bio->bi_sector - tmp_dev->start_sector
		+ tmp_dev->rdev->data_offset;
L
Linus Torvalds 已提交
377 378 379 380 381 382 383 384 385 386 387

	return 1;
}

static void linear_status (struct seq_file *seq, mddev_t *mddev)
{

	seq_printf(seq, " %dk rounding", mddev->chunk_size/1024);
}


388
static struct mdk_personality linear_personality =
L
Linus Torvalds 已提交
389 390
{
	.name		= "linear",
391
	.level		= LEVEL_LINEAR,
L
Linus Torvalds 已提交
392 393 394 395 396
	.owner		= THIS_MODULE,
	.make_request	= linear_make_request,
	.run		= linear_run,
	.stop		= linear_stop,
	.status		= linear_status,
397
	.hot_add_disk	= linear_add,
398
	.size		= linear_size,
L
Linus Torvalds 已提交
399 400 401 402
};

static int __init linear_init (void)
{
403
	return register_md_personality (&linear_personality);
L
Linus Torvalds 已提交
404 405 406 407
}

static void linear_exit (void)
{
408
	unregister_md_personality (&linear_personality);
L
Linus Torvalds 已提交
409 410 411 412 413 414
}


module_init(linear_init);
module_exit(linear_exit);
MODULE_LICENSE("GPL");
415 416
MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/
MODULE_ALIAS("md-linear");
417
MODULE_ALIAS("md-level--1");