multipath.c 14.1 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 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
/*
 * multipath.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
 *
 * Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
 *
 * MULTIPATH management functions.
 *
 * derived from raid1.c.
 *
 * 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/slab.h>
#include <linux/spinlock.h>
#include <linux/raid/multipath.h>
#include <linux/buffer_head.h>
#include <asm/atomic.h>

#define MAJOR_NR MD_MAJOR
#define MD_DRIVER
#define MD_PERSONALITY

#define MAX_WORK_PER_DISK 128

#define	NR_RESERVED_BUFS	32


static int multipath_map (multipath_conf_t *conf)
{
	int i, disks = conf->raid_disks;

	/*
	 * Later we do read balancing on the read side 
	 * now we use the first available disk.
	 */

	rcu_read_lock();
	for (i = 0; i < disks; i++) {
49
		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
50
		if (rdev && test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();
			return i;
		}
	}
	rcu_read_unlock();

	printk(KERN_ERR "multipath_map(): no more operational IO paths?\n");
	return (-1);
}

static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
{
	unsigned long flags;
	mddev_t *mddev = mp_bh->mddev;
	multipath_conf_t *conf = mddev_to_conf(mddev);

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&mp_bh->retry_list, &conf->retry_list);
	spin_unlock_irqrestore(&conf->device_lock, flags);
	md_wakeup_thread(mddev->thread);
}


/*
 * multipath_end_bh_io() is called when we have finished servicing a multipathed
 * operation and are ready to return a success/failure code to the buffer
 * cache layer.
 */
static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
{
	struct bio *bio = mp_bh->master_bio;
	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);

85
	bio_endio(bio, err);
L
Linus Torvalds 已提交
86 87 88
	mempool_free(mp_bh, conf->pool);
}

89
static void multipath_end_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);
	mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev;

	if (uptodate)
		multipath_end_bh_io(mp_bh, 0);
	else if (!bio_rw_ahead(bio)) {
		/*
		 * oops, IO error:
		 */
		char b[BDEVNAME_SIZE];
		md_error (mp_bh->mddev, rdev);
		printk(KERN_ERR "multipath: %s: rescheduling sector %llu\n", 
		       bdevname(rdev->bdev,b), 
		       (unsigned long long)bio->bi_sector);
		multipath_reschedule_retry(mp_bh);
	} else
		multipath_end_bh_io(mp_bh, error);
	rdev_dec_pending(rdev, conf->mddev);
}

static void unplug_slaves(mddev_t *mddev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	int i;

	rcu_read_lock();
	for (i=0; i<mddev->raid_disks; i++) {
120
		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
121 122
		if (rdev && !test_bit(Faulty, &rdev->flags)
		    && atomic_read(&rdev->nr_pending)) {
123
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
124 125 126 127

			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();

128
			blk_unplug(r_queue);
L
Linus Torvalds 已提交
129 130 131 132 133 134 135 136

			rdev_dec_pending(rdev, mddev);
			rcu_read_lock();
		}
	}
	rcu_read_unlock();
}

137
static void multipath_unplug(struct request_queue *q)
L
Linus Torvalds 已提交
138 139 140 141 142
{
	unplug_slaves(q->queuedata);
}


143
static int multipath_make_request (struct request_queue *q, struct bio * bio)
L
Linus Torvalds 已提交
144 145 146 147 148
{
	mddev_t *mddev = q->queuedata;
	multipath_conf_t *conf = mddev_to_conf(mddev);
	struct multipath_bh * mp_bh;
	struct multipath_info *multipath;
149
	const int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
150

151
	if (unlikely(bio_barrier(bio))) {
152
		bio_endio(bio, -EOPNOTSUPP);
153 154 155
		return 0;
	}

L
Linus Torvalds 已提交
156 157 158 159 160
	mp_bh = mempool_alloc(conf->pool, GFP_NOIO);

	mp_bh->master_bio = bio;
	mp_bh->mddev = mddev;

161 162
	disk_stat_inc(mddev->gendisk, ios[rw]);
	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
L
Linus Torvalds 已提交
163 164 165

	mp_bh->path = multipath_map(conf);
	if (mp_bh->path < 0) {
166
		bio_endio(bio, -EIO);
L
Linus Torvalds 已提交
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
		mempool_free(mp_bh, conf->pool);
		return 0;
	}
	multipath = conf->multipaths + mp_bh->path;

	mp_bh->bio = *bio;
	mp_bh->bio.bi_sector += multipath->rdev->data_offset;
	mp_bh->bio.bi_bdev = multipath->rdev->bdev;
	mp_bh->bio.bi_rw |= (1 << BIO_RW_FAILFAST);
	mp_bh->bio.bi_end_io = multipath_end_request;
	mp_bh->bio.bi_private = mp_bh;
	generic_make_request(&mp_bh->bio);
	return 0;
}

static void multipath_status (struct seq_file *seq, mddev_t *mddev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	int i;
	
	seq_printf (seq, " [%d/%d] [", conf->raid_disks,
						 conf->working_disks);
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->multipaths[i].rdev && 
192
			       test_bit(In_sync, &conf->multipaths[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
193 194 195
	seq_printf (seq, "]");
}

196 197 198 199 200 201 202 203 204 205
static int multipath_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	multipath_conf_t *conf = mddev_to_conf(mddev);
	int i, ret = 0;

	rcu_read_lock();
	for (i = 0; i < mddev->raid_disks ; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
206
			struct request_queue *q = bdev_get_queue(rdev->bdev);
207 208 209 210 211 212 213 214 215 216 217

			ret |= bdi_congested(&q->backing_dev_info, bits);
			/* Just like multipath_map, we just check the
			 * first available device
			 */
			break;
		}
	}
	rcu_read_unlock();
	return ret;
}
L
Linus Torvalds 已提交
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238

/*
 * Careful, this can execute in IRQ contexts as well!
 */
static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);

	if (conf->working_disks <= 1) {
		/*
		 * Uh oh, we can do nothing if this is our last path, but
		 * first check if this is a queued request for a device
		 * which has just failed.
		 */
		printk(KERN_ALERT 
			"multipath: only one IO path left and IO error.\n");
		/* leave it active... it's all we have */
	} else {
		/*
		 * Mark disk as unusable
		 */
239
		if (!test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
240
			char b[BDEVNAME_SIZE];
241 242
			clear_bit(In_sync, &rdev->flags);
			set_bit(Faulty, &rdev->flags);
243
			set_bit(MD_CHANGE_DEVS, &mddev->flags);
L
Linus Torvalds 已提交
244
			conf->working_disks--;
245
			mddev->degraded++;
L
Linus Torvalds 已提交
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
			printk(KERN_ALERT "multipath: IO failure on %s,"
				" disabling IO path. \n	Operation continuing"
				" on %d IO paths.\n",
				bdevname (rdev->bdev,b),
				conf->working_disks);
		}
	}
}

static void print_multipath_conf (multipath_conf_t *conf)
{
	int i;
	struct multipath_info *tmp;

	printk("MULTIPATH conf printout:\n");
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
	printk(" --- wd:%d rd:%d\n", conf->working_disks,
			 conf->raid_disks);

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->multipaths + i;
		if (tmp->rdev)
			printk(" disk%d, o:%d, dev:%s\n",
273
				i,!test_bit(Faulty, &tmp->rdev->flags),
L
Linus Torvalds 已提交
274 275 276 277 278 279 280 281
			       bdevname(tmp->rdev->bdev,b));
	}
}


static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	multipath_conf_t *conf = mddev->private;
282
	struct request_queue *q;
L
Linus Torvalds 已提交
283 284 285 286 287 288 289 290
	int found = 0;
	int path;
	struct multipath_info *p;

	print_multipath_conf(conf);

	for (path=0; path<mddev->raid_disks; path++) 
		if ((p=conf->multipaths+path)->rdev == NULL) {
291 292
			q = rdev->bdev->bd_disk->queue;
			blk_queue_stack_limits(mddev->queue, q);
L
Linus Torvalds 已提交
293 294 295 296 297 298 299

		/* 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.
		 * (Note: it is very unlikely that a device with
		 * merge_bvec_fn will be involved in multipath.)
		 */
300
			if (q->merge_bvec_fn &&
L
Linus Torvalds 已提交
301 302 303 304
			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
				blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

			conf->working_disks++;
305
			mddev->degraded--;
L
Linus Torvalds 已提交
306
			rdev->raid_disk = path;
307
			set_bit(In_sync, &rdev->flags);
308
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
			found = 1;
		}

	print_multipath_conf(conf);
	return found;
}

static int multipath_remove_disk(mddev_t *mddev, int number)
{
	multipath_conf_t *conf = mddev->private;
	int err = 0;
	mdk_rdev_t *rdev;
	struct multipath_info *p = conf->multipaths + number;

	print_multipath_conf(conf);

	rdev = p->rdev;
	if (rdev) {
327
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
328 329 330 331 332 333
		    atomic_read(&rdev->nr_pending)) {
			printk(KERN_ERR "hot-remove-disk, slot %d is identified"				" but is still operational!\n", number);
			err = -EBUSY;
			goto abort;
		}
		p->rdev = NULL;
334
		synchronize_rcu();
L
Linus Torvalds 已提交
335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_multipath_conf(conf);
	return err;
}



/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working multipaths.
 *	2.	Updates the raid superblock when problems encounter.
 *	3.	Performs writes following reads for array syncronising.
 */

static void multipathd (mddev_t *mddev)
{
	struct multipath_bh *mp_bh;
	struct bio *bio;
	unsigned long flags;
	multipath_conf_t *conf = mddev_to_conf(mddev);
	struct list_head *head = &conf->retry_list;

	md_check_recovery(mddev);
	for (;;) {
		char b[BDEVNAME_SIZE];
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head))
			break;
		mp_bh = list_entry(head->prev, struct multipath_bh, retry_list);
		list_del(head->prev);
		spin_unlock_irqrestore(&conf->device_lock, flags);

		bio = &mp_bh->bio;
		bio->bi_sector = mp_bh->master_bio->bi_sector;
		
		if ((mp_bh->path = multipath_map (conf))<0) {
			printk(KERN_ALERT "multipath: %s: unrecoverable IO read"
				" error for block %llu\n",
				bdevname(bio->bi_bdev,b),
				(unsigned long long)bio->bi_sector);
			multipath_end_bh_io(mp_bh, -EIO);
		} else {
			printk(KERN_ERR "multipath: %s: redirecting sector %llu"
				" to another IO path\n",
				bdevname(bio->bi_bdev,b),
				(unsigned long long)bio->bi_sector);
			*bio = *(mp_bh->master_bio);
			bio->bi_sector += conf->multipaths[mp_bh->path].rdev->data_offset;
			bio->bi_bdev = conf->multipaths[mp_bh->path].rdev->bdev;
			bio->bi_rw |= (1 << BIO_RW_FAILFAST);
			bio->bi_end_io = multipath_end_request;
			bio->bi_private = mp_bh;
			generic_make_request(bio);
		}
	}
	spin_unlock_irqrestore(&conf->device_lock, flags);
}

static int multipath_run (mddev_t *mddev)
{
	multipath_conf_t *conf;
	int disk_idx;
	struct multipath_info *disk;
	mdk_rdev_t *rdev;
	struct list_head *tmp;

	if (mddev->level != LEVEL_MULTIPATH) {
		printk("multipath: %s: raid level not set to multipath IO (%d)\n",
		       mdname(mddev), mddev->level);
		goto out;
	}
	/*
	 * copy the already verified devices into our private MULTIPATH
	 * bookkeeping area. [whatever we allocate in multipath_run(),
	 * should be freed in multipath_stop()]
	 */

420
	conf = kzalloc(sizeof(multipath_conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428
	mddev->private = conf;
	if (!conf) {
		printk(KERN_ERR 
			"multipath: couldn't allocate memory for %s\n",
			mdname(mddev));
		goto out;
	}

429
	conf->multipaths = kzalloc(sizeof(struct multipath_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
430 431 432 433 434 435 436 437 438
				   GFP_KERNEL);
	if (!conf->multipaths) {
		printk(KERN_ERR 
			"multipath: couldn't allocate memory for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}

	conf->working_disks = 0;
439
	rdev_for_each(rdev, tmp, mddev) {
L
Linus Torvalds 已提交
440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
		disk_idx = rdev->raid_disk;
		if (disk_idx < 0 ||
		    disk_idx >= mddev->raid_disks)
			continue;

		disk = conf->multipaths + disk_idx;
		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, not that we ever expect a device with
		 * a merge_bvec_fn to be involved in multipath */
		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);

457
		if (!test_bit(Faulty, &rdev->flags))
L
Linus Torvalds 已提交
458 459 460 461 462 463 464 465 466 467 468 469 470
			conf->working_disks++;
	}

	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	spin_lock_init(&conf->device_lock);
	INIT_LIST_HEAD(&conf->retry_list);

	if (!conf->working_disks) {
		printk(KERN_ERR "multipath: no operational IO paths for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}
471
	mddev->degraded = conf->raid_disks - conf->working_disks;
L
Linus Torvalds 已提交
472

473 474
	conf->pool = mempool_create_kzalloc_pool(NR_RESERVED_BUFS,
						 sizeof(struct multipath_bh));
L
Linus Torvalds 已提交
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
	if (conf->pool == NULL) {
		printk(KERN_ERR 
			"multipath: couldn't allocate memory for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}

	{
		mddev->thread = md_register_thread(multipathd, mddev, "%s_multipath");
		if (!mddev->thread) {
			printk(KERN_ERR "multipath: couldn't allocate thread"
				" for %s\n", mdname(mddev));
			goto out_free_conf;
		}
	}

	printk(KERN_INFO 
		"multipath: array %s active with %d out of %d IO paths\n",
		mdname(mddev), conf->working_disks, mddev->raid_disks);
	/*
	 * Ok, everything is just fine now
	 */
	mddev->array_size = mddev->size;
498 499

	mddev->queue->unplug_fn = multipath_unplug;
500 501
	mddev->queue->backing_dev_info.congested_fn = multipath_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
502

L
Linus Torvalds 已提交
503 504 505 506 507
	return 0;

out_free_conf:
	if (conf->pool)
		mempool_destroy(conf->pool);
508
	kfree(conf->multipaths);
L
Linus Torvalds 已提交
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529
	kfree(conf);
	mddev->private = NULL;
out:
	return -EIO;
}


static int multipath_stop (mddev_t *mddev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);

	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
	mempool_destroy(conf->pool);
	kfree(conf->multipaths);
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

530
static struct mdk_personality multipath_personality =
L
Linus Torvalds 已提交
531 532
{
	.name		= "multipath",
533
	.level		= LEVEL_MULTIPATH,
L
Linus Torvalds 已提交
534 535 536 537 538 539 540 541 542 543 544 545
	.owner		= THIS_MODULE,
	.make_request	= multipath_make_request,
	.run		= multipath_run,
	.stop		= multipath_stop,
	.status		= multipath_status,
	.error_handler	= multipath_error,
	.hot_add_disk	= multipath_add_disk,
	.hot_remove_disk= multipath_remove_disk,
};

static int __init multipath_init (void)
{
546
	return register_md_personality (&multipath_personality);
L
Linus Torvalds 已提交
547 548 549 550
}

static void __exit multipath_exit (void)
{
551
	unregister_md_personality (&multipath_personality);
L
Linus Torvalds 已提交
552 553 554 555 556 557
}

module_init(multipath_init);
module_exit(multipath_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-7"); /* MULTIPATH */
558
MODULE_ALIAS("md-multipath");
559
MODULE_ALIAS("md-level--4");