[linux.git] / drivers / md / md-multipath.c

/*
 * 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/blkdev.h>
#include <linux/module.h>
#include <linux/raid/md_u.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
#include "md.h"
#include "md-multipath.h"

#define MAX_WORK_PER_DISK 128

#define	NR_RESERVED_BUFS	32

static int multipath_map (struct mpconf *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++) {
		struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
		if (rdev && test_bit(In_sync, &rdev->flags) &&
		    !test_bit(Faulty, &rdev->flags)) {
			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();
			return i;
		}
	}
	rcu_read_unlock();

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

static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
{
	unsigned long flags;
	struct mddev *mddev = mp_bh->mddev;
	struct mpconf *conf = mddev->private;

	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, blk_status_t status)
{
	struct bio *bio = mp_bh->master_bio;
	struct mpconf *conf = mp_bh->mddev->private;

	bio->bi_status = status;
	bio_endio(bio);
	mempool_free(mp_bh, &conf->pool);
}

static void multipath_end_request(struct bio *bio)
{
	struct multipath_bh *mp_bh = bio->bi_private;
	struct mpconf *conf = mp_bh->mddev->private;
	struct md_rdev *rdev = conf->multipaths[mp_bh->path].rdev;

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

static bool multipath_make_request(struct mddev *mddev, struct bio * bio)
{
	struct mpconf *conf = mddev->private;
	struct multipath_bh * mp_bh;
	struct multipath_info *multipath;

	if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
		md_flush_request(mddev, bio);
		return true;
	}

	mp_bh = mempool_alloc(&conf->pool, GFP_NOIO);

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

	mp_bh->path = multipath_map(conf);
	if (mp_bh->path < 0) {
		bio_io_error(bio);
		mempool_free(mp_bh, &conf->pool);
		return true;
	}
	multipath = conf->multipaths + mp_bh->path;

	bio_init(&mp_bh->bio, NULL, 0);
	__bio_clone_fast(&mp_bh->bio, bio);

	mp_bh->bio.bi_iter.bi_sector += multipath->rdev->data_offset;
	bio_set_dev(&mp_bh->bio, multipath->rdev->bdev);
	mp_bh->bio.bi_opf |= REQ_FAILFAST_TRANSPORT;
	mp_bh->bio.bi_end_io = multipath_end_request;
	mp_bh->bio.bi_private = mp_bh;
	mddev_check_writesame(mddev, &mp_bh->bio);
	mddev_check_write_zeroes(mddev, &mp_bh->bio);
	generic_make_request(&mp_bh->bio);
	return true;
}

static void multipath_status(struct seq_file *seq, struct mddev *mddev)
{
	struct mpconf *conf = mddev->private;
	int i;

	seq_printf (seq, " [%d/%d] [", conf->raid_disks,
		    conf->raid_disks - mddev->degraded);
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
		struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
		seq_printf (seq, "%s", rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
	seq_putc(seq, ']');
}

static int multipath_congested(struct mddev *mddev, int bits)
{
	struct mpconf *conf = mddev->private;
	int i, ret = 0;

	rcu_read_lock();
	for (i = 0; i < mddev->raid_disks ; i++) {
		struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
			struct request_queue *q = bdev_get_queue(rdev->bdev);

			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;
}

/*
 * Careful, this can execute in IRQ contexts as well!
 */
static void multipath_error (struct mddev *mddev, struct md_rdev *rdev)
{
	struct mpconf *conf = mddev->private;
	char b[BDEVNAME_SIZE];

	if (conf->raid_disks - mddev->degraded <= 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.
		 */
		pr_warn("multipath: only one IO path left and IO error.\n");
		/* leave it active... it's all we have */
		return;
	}
	/*
	 * Mark disk as unusable
	 */
	if (test_and_clear_bit(In_sync, &rdev->flags)) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		mddev->degraded++;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
	set_bit(Faulty, &rdev->flags);
	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
	pr_err("multipath: IO failure on %s, disabling IO path.\n"
	       "multipath: Operation continuing on %d IO paths.\n",
	       bdevname(rdev->bdev, b),
	       conf->raid_disks - mddev->degraded);
}

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

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

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->multipaths + i;
		if (tmp->rdev)
			pr_debug(" disk%d, o:%d, dev:%s\n",
				 i,!test_bit(Faulty, &tmp->rdev->flags),
				 bdevname(tmp->rdev->bdev,b));
	}
}

static int multipath_add_disk(struct mddev *mddev, struct md_rdev *rdev)
{
	struct mpconf *conf = mddev->private;
	int err = -EEXIST;
	int path;
	struct multipath_info *p;
	int first = 0;
	int last = mddev->raid_disks - 1;

	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	print_multipath_conf(conf);

	for (path = first; path <= last; path++)
		if ((p=conf->multipaths+path)->rdev == NULL) {
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);

			err = md_integrity_add_rdev(rdev, mddev);
			if (err)
				break;
			spin_lock_irq(&conf->device_lock);
			mddev->degraded--;
			rdev->raid_disk = path;
			set_bit(In_sync, &rdev->flags);
			spin_unlock_irq(&conf->device_lock);
			rcu_assign_pointer(p->rdev, rdev);
			err = 0;
			break;
		}

	print_multipath_conf(conf);

	return err;
}

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

	print_multipath_conf(conf);

	if (rdev == p->rdev) {
		if (test_bit(In_sync, &rdev->flags) ||
		    atomic_read(&rdev->nr_pending)) {
			pr_warn("hot-remove-disk, slot %d is identified but is still operational!\n", number);
			err = -EBUSY;
			goto abort;
		}
		p->rdev = NULL;
		if (!test_bit(RemoveSynchronized, &rdev->flags)) {
			synchronize_rcu();
			if (atomic_read(&rdev->nr_pending)) {
				/* lost the race, try later */
				err = -EBUSY;
				p->rdev = rdev;
				goto abort;
			}
		}
		err = md_integrity_register(mddev);
	}
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(struct md_thread *thread)
{
	struct mddev *mddev = thread->mddev;
	struct multipath_bh *mp_bh;
	struct bio *bio;
	unsigned long flags;
	struct mpconf *conf = mddev->private;
	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_iter.bi_sector = mp_bh->master_bio->bi_iter.bi_sector;

		if ((mp_bh->path = multipath_map (conf))<0) {
			pr_err("multipath: %s: unrecoverable IO read error for block %llu\n",
			       bio_devname(bio, b),
			       (unsigned long long)bio->bi_iter.bi_sector);
			multipath_end_bh_io(mp_bh, BLK_STS_IOERR);
		} else {
			pr_err("multipath: %s: redirecting sector %llu to another IO path\n",
			       bio_devname(bio, b),
			       (unsigned long long)bio->bi_iter.bi_sector);
			*bio = *(mp_bh->master_bio);
			bio->bi_iter.bi_sector +=
				conf->multipaths[mp_bh->path].rdev->data_offset;
			bio_set_dev(bio, conf->multipaths[mp_bh->path].rdev->bdev);
			bio->bi_opf |= REQ_FAILFAST_TRANSPORT;
			bio->bi_end_io = multipath_end_request;
			bio->bi_private = mp_bh;
			generic_make_request(bio);
		}
	}
	spin_unlock_irqrestore(&conf->device_lock, flags);
}

static sector_t multipath_size(struct mddev *mddev, sector_t sectors, int raid_disks)
{
	WARN_ONCE(sectors || raid_disks,
		  "%s does not support generic reshape\n", __func__);

	return mddev->dev_sectors;
}

static int multipath_run (struct mddev *mddev)
{
	struct mpconf *conf;
	int disk_idx;
	struct multipath_info *disk;
	struct md_rdev *rdev;
	int working_disks;
	int ret;

	if (md_check_no_bitmap(mddev))
		return -EINVAL;

	if (mddev->level != LEVEL_MULTIPATH) {
		pr_warn("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_free()]
	 */

	conf = kzalloc(sizeof(struct mpconf), GFP_KERNEL);
	mddev->private = conf;
	if (!conf)
		goto out;

	conf->multipaths = kcalloc(mddev->raid_disks,
				   sizeof(struct multipath_info),
				   GFP_KERNEL);
	if (!conf->multipaths)
		goto out_free_conf;

	working_disks = 0;
	rdev_for_each(rdev, mddev) {
		disk_idx = rdev->raid_disk;
		if (disk_idx < 0 ||
		    disk_idx >= mddev->raid_disks)
			continue;

		disk = conf->multipaths + disk_idx;
		disk->rdev = rdev;
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);

		if (!test_bit(Faulty, &rdev->flags))
			working_disks++;
	}

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

	if (!working_disks) {
		pr_warn("multipath: no operational IO paths for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}
	mddev->degraded = conf->raid_disks - working_disks;

	ret = mempool_init_kmalloc_pool(&conf->pool, NR_RESERVED_BUFS,
					sizeof(struct multipath_bh));
	if (ret)
		goto out_free_conf;

	mddev->thread = md_register_thread(multipathd, mddev,
					   "multipath");
	if (!mddev->thread)
		goto out_free_conf;

	pr_info("multipath: array %s active with %d out of %d IO paths\n",
		mdname(mddev), conf->raid_disks - mddev->degraded,
		mddev->raid_disks);
	/*
	 * Ok, everything is just fine now
	 */
	md_set_array_sectors(mddev, multipath_size(mddev, 0, 0));

	if (md_integrity_register(mddev))
		goto out_free_conf;

	return 0;

out_free_conf:
	mempool_exit(&conf->pool);
	kfree(conf->multipaths);
	kfree(conf);
	mddev->private = NULL;
out:
	return -EIO;
}

static void multipath_free(struct mddev *mddev, void *priv)
{
	struct mpconf *conf = priv;

	mempool_exit(&conf->pool);
	kfree(conf->multipaths);
	kfree(conf);
}

static struct md_personality multipath_personality =
{
	.name		= "multipath",
	.level		= LEVEL_MULTIPATH,
	.owner		= THIS_MODULE,
	.make_request	= multipath_make_request,
	.run		= multipath_run,
	.free		= multipath_free,
	.status		= multipath_status,
	.error_handler	= multipath_error,
	.hot_add_disk	= multipath_add_disk,
	.hot_remove_disk= multipath_remove_disk,
	.size		= multipath_size,
	.congested	= multipath_congested,
};

static int __init multipath_init (void)
{
	return register_md_personality (&multipath_personality);
}

static void __exit multipath_exit (void)
{
	unregister_md_personality (&multipath_personality);
}

module_init(multipath_init);
module_exit(multipath_exit);
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("simple multi-path personality for MD");
MODULE_ALIAS("md-personality-7"); /* MULTIPATH */
MODULE_ALIAS("md-multipath");
MODULE_ALIAS("md-level--4");
Commit	Line	Data
1da177e4 LT	1	/*
	2	* multipath.c : Multiple Devices driver for Linux
	3	*
	4	* Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
	5	*
	6	* Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
	7	*
	8	* MULTIPATH management functions.
	9	*
	10	* derived from raid1.c.
	11	*
	12	* This program is free software; you can redistribute it and/or modify
	13	* it under the terms of the GNU General Public License as published by
	14	* the Free Software Foundation; either version 2, or (at your option)
	15	* any later version.
	16	*
	17	* You should have received a copy of the GNU General Public License
	18	* (for example /usr/src/linux/COPYING); if not, write to the Free
	19	* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
	20	*/
	21
bff61975	22	#include <linux/blkdev.h>
056075c7	23	#include <linux/module.h>
bff61975	24	#include <linux/raid/md_u.h>
bff61975	25	#include <linux/seq_file.h>
5a0e3ad6	26	#include <linux/slab.h>
43b2e5d8	27	#include "md.h"
935fe098	28	#include "md-multipath.h"
1da177e4 LT	29
	30	#define MAX_WORK_PER_DISK 128
	31
	32	#define NR_RESERVED_BUFS 32
	33
69724e28	34	static int multipath_map (struct mpconf *conf)
1da177e4 LT	35	{
	36	int i, disks = conf->raid_disks;
	37
	38	/*
f72ffdd6	39	* Later we do read balancing on the read side
1da177e4 LT	40	* now we use the first available disk.
	41	*/
	42
	43	rcu_read_lock();
	44	for (i = 0; i < disks; i++) {
3cb03002	45	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
f5b67ae8 N	46	if (rdev && test_bit(In_sync, &rdev->flags) &&
f5b67ae8 N	47	!test_bit(Faulty, &rdev->flags)) {
1da177e4 LT	48	atomic_inc(&rdev->nr_pending);
	49	rcu_read_unlock();
	50	return i;
	51	}
	52	}
	53	rcu_read_unlock();
	54
7279694d	55	pr_crit_ratelimited("multipath_map(): no more operational IO paths?\n");
1da177e4 LT	56	return (-1);
	57	}
	58
	59	static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
	60	{
	61	unsigned long flags;
fd01b88c	62	struct mddev *mddev = mp_bh->mddev;
69724e28	63	struct mpconf *conf = mddev->private;
1da177e4 LT	64
	65	spin_lock_irqsave(&conf->device_lock, flags);
	66	list_add(&mp_bh->retry_list, &conf->retry_list);
	67	spin_unlock_irqrestore(&conf->device_lock, flags);
	68	md_wakeup_thread(mddev->thread);
	69	}
	70
1da177e4 LT	71	/*
	72	* multipath_end_bh_io() is called when we have finished servicing a multipathed
	73	* operation and are ready to return a success/failure code to the buffer
	74	* cache layer.
	75	*/
4e4cbee9	76	static void multipath_end_bh_io(struct multipath_bh *mp_bh, blk_status_t status)
1da177e4 LT	77	{
1da177e4 LT	78	struct bio *bio = mp_bh->master_bio;
69724e28	79	struct mpconf *conf = mp_bh->mddev->private;
1da177e4	80
4e4cbee9	81	bio->bi_status = status;
4246a0b6	82	bio_endio(bio);
afeee514	83	mempool_free(mp_bh, &conf->pool);
1da177e4 LT	84	}
1da177e4 LT	85
4246a0b6	86	static void multipath_end_request(struct bio *bio)
1da177e4	87	{
7b92813c	88	struct multipath_bh *mp_bh = bio->bi_private;
69724e28	89	struct mpconf *conf = mp_bh->mddev->private;
3cb03002	90	struct md_rdev *rdev = conf->multipaths[mp_bh->path].rdev;
1da177e4	91
4e4cbee9	92	if (!bio->bi_status)
1da177e4	93	multipath_end_bh_io(mp_bh, 0);
1eff9d32	94	else if (!(bio->bi_opf & REQ_RAHEAD)) {
1da177e4 LT	95	/*
	96	* oops, IO error:
	97	*/
	98	char b[BDEVNAME_SIZE];
	99	md_error (mp_bh->mddev, rdev);
7279694d N	100	pr_info("multipath: %s: rescheduling sector %llu\n",
	101	bdevname(rdev->bdev,b),
	102	(unsigned long long)bio->bi_iter.bi_sector);
1da177e4 LT	103	multipath_reschedule_retry(mp_bh);
1da177e4 LT	104	} else
4e4cbee9	105	multipath_end_bh_io(mp_bh, bio->bi_status);
1da177e4	106	rdev_dec_pending(rdev, conf->mddev);
1da177e4 LT	107	}
1da177e4 LT	108
cc27b0c7	109	static bool multipath_make_request(struct mddev mddev, struct bio bio)
1da177e4	110	{
69724e28	111	struct mpconf *conf = mddev->private;
1da177e4 LT	112	struct multipath_bh * mp_bh;
	113	struct multipath_info *multipath;
	114
1eff9d32	115	if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
e9c7469b	116	md_flush_request(mddev, bio);
cc27b0c7	117	return true;
e5dcdd80 N	118	}
e5dcdd80 N	119
afeee514	120	mp_bh = mempool_alloc(&conf->pool, GFP_NOIO);
1da177e4 LT	121
	122	mp_bh->master_bio = bio;
	123	mp_bh->mddev = mddev;
	124
1da177e4 LT	125	mp_bh->path = multipath_map(conf);
1da177e4 LT	126	if (mp_bh->path < 0) {
4246a0b6	127	bio_io_error(bio);
afeee514	128	mempool_free(mp_bh, &conf->pool);
cc27b0c7	129	return true;
1da177e4 LT	130	}
	131	multipath = conf->multipaths + mp_bh->path;
	132
3a83f467	133	bio_init(&mp_bh->bio, NULL, 0);
fafcde3a ML	134	__bio_clone_fast(&mp_bh->bio, bio);
fafcde3a ML	135
4f024f37	136	mp_bh->bio.bi_iter.bi_sector += multipath->rdev->data_offset;
74d46992	137	bio_set_dev(&mp_bh->bio, multipath->rdev->bdev);
1eff9d32	138	mp_bh->bio.bi_opf \|= REQ_FAILFAST_TRANSPORT;
1da177e4 LT	139	mp_bh->bio.bi_end_io = multipath_end_request;
1da177e4 LT	140	mp_bh->bio.bi_private = mp_bh;
26483819	141	mddev_check_writesame(mddev, &mp_bh->bio);
3deff1a7	142	mddev_check_write_zeroes(mddev, &mp_bh->bio);
1da177e4	143	generic_make_request(&mp_bh->bio);
cc27b0c7	144	return true;
1da177e4 LT	145	}
1da177e4 LT	146
40cf2123	147	static void multipath_status(struct seq_file seq, struct mddev mddev)
1da177e4	148	{
69724e28	149	struct mpconf *conf = mddev->private;
1da177e4	150	int i;
f72ffdd6	151
1da177e4	152	seq_printf (seq, " [%d/%d] [", conf->raid_disks,
92f861a7	153	conf->raid_disks - mddev->degraded);
40cf2123 N	154	rcu_read_lock();
	155	for (i = 0; i < conf->raid_disks; i++) {
	156	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
	157	seq_printf (seq, "%s", rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	158	}
	159	rcu_read_unlock();
3acdb7b5	160	seq_putc(seq, ']');
1da177e4 LT	161	}
1da177e4 LT	162
5c675f83	163	static int multipath_congested(struct mddev *mddev, int bits)
0d129228	164	{
69724e28	165	struct mpconf *conf = mddev->private;
0d129228 N	166	int i, ret = 0;
	167
	168	rcu_read_lock();
	169	for (i = 0; i < mddev->raid_disks ; i++) {
3cb03002	170	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
0d129228	171	if (rdev && !test_bit(Faulty, &rdev->flags)) {
165125e1	172	struct request_queue *q = bdev_get_queue(rdev->bdev);
0d129228	173
dc3b17cc	174	ret \|= bdi_congested(q->backing_dev_info, bits);
0d129228 N	175	/* Just like multipath_map, we just check the
	176	* first available device
	177	*/
	178	break;
	179	}
	180	}
	181	rcu_read_unlock();
	182	return ret;
	183	}
1da177e4 LT	184
	185	/*
	186	* Careful, this can execute in IRQ contexts as well!
	187	*/
fd01b88c	188	static void multipath_error (struct mddev mddev, struct md_rdev rdev)
1da177e4	189	{
69724e28	190	struct mpconf *conf = mddev->private;
6f8d0c77	191	char b[BDEVNAME_SIZE];
1da177e4	192
92f861a7	193	if (conf->raid_disks - mddev->degraded <= 1) {
1da177e4 LT	194	/*
	195	* Uh oh, we can do nothing if this is our last path, but
	196	* first check if this is a queued request for a device
	197	* which has just failed.
	198	*/
7279694d	199	pr_warn("multipath: only one IO path left and IO error.\n");
1da177e4	200	/* leave it active... it's all we have */
6f8d0c77 N	201	return;
	202	}
	203	/*
	204	* Mark disk as unusable
	205	*/
	206	if (test_and_clear_bit(In_sync, &rdev->flags)) {
	207	unsigned long flags;
	208	spin_lock_irqsave(&conf->device_lock, flags);
	209	mddev->degraded++;
	210	spin_unlock_irqrestore(&conf->device_lock, flags);
1da177e4	211	}
6f8d0c77	212	set_bit(Faulty, &rdev->flags);
2953079c	213	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7279694d N	214	pr_err("multipath: IO failure on %s, disabling IO path.\n"
7279694d N	215	"multipath: Operation continuing on %d IO paths.\n",
6f8d0c77 N	216	bdevname(rdev->bdev, b),
6f8d0c77 N	217	conf->raid_disks - mddev->degraded);
1da177e4 LT	218	}
1da177e4 LT	219
69724e28	220	static void print_multipath_conf (struct mpconf *conf)
1da177e4 LT	221	{
	222	int i;
	223	struct multipath_info *tmp;
	224
7279694d	225	pr_debug("MULTIPATH conf printout:\n");
1da177e4	226	if (!conf) {
7279694d	227	pr_debug("(conf==NULL)\n");
1da177e4 LT	228	return;
1da177e4 LT	229	}
7279694d N	230	pr_debug(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
7279694d N	231	conf->raid_disks);
1da177e4 LT	232
	233	for (i = 0; i < conf->raid_disks; i++) {
	234	char b[BDEVNAME_SIZE];
	235	tmp = conf->multipaths + i;
	236	if (tmp->rdev)
7279694d N	237	pr_debug(" disk%d, o:%d, dev:%s\n",
	238	i,!test_bit(Faulty, &tmp->rdev->flags),
	239	bdevname(tmp->rdev->bdev,b));
1da177e4 LT	240	}
	241	}
	242
fd01b88c	243	static int multipath_add_disk(struct mddev mddev, struct md_rdev rdev)
1da177e4	244	{
69724e28	245	struct mpconf *conf = mddev->private;
199050ea	246	int err = -EEXIST;
1da177e4 LT	247	int path;
1da177e4 LT	248	struct multipath_info *p;
6c2fce2e NB	249	int first = 0;
	250	int last = mddev->raid_disks - 1;
	251
	252	if (rdev->raid_disk >= 0)
	253	first = last = rdev->raid_disk;
1da177e4 LT	254
	255	print_multipath_conf(conf);
	256
6c2fce2e	257	for (path = first; path <= last; path++)
1da177e4	258	if ((p=conf->multipaths+path)->rdev == NULL) {
8f6c2e4b MP	259	disk_stack_limits(mddev->gendisk, rdev->bdev,
8f6c2e4b MP	260	rdev->data_offset << 9);
1da177e4	261
1501efad DW	262	err = md_integrity_add_rdev(rdev, mddev);
	263	if (err)
	264	break;
6f8d0c77	265	spin_lock_irq(&conf->device_lock);
750a8f3e	266	mddev->degraded--;
1da177e4	267	rdev->raid_disk = path;
b2d444d7	268	set_bit(In_sync, &rdev->flags);
6f8d0c77	269	spin_unlock_irq(&conf->device_lock);
d6065f7b	270	rcu_assign_pointer(p->rdev, rdev);
199050ea NB	271	err = 0;
199050ea NB	272	break;
1da177e4 LT	273	}
	274
	275	print_multipath_conf(conf);
199050ea NB	276
199050ea NB	277	return err;
1da177e4 LT	278	}
1da177e4 LT	279
b8321b68	280	static int multipath_remove_disk(struct mddev mddev, struct md_rdev rdev)
1da177e4	281	{
69724e28	282	struct mpconf *conf = mddev->private;
1da177e4	283	int err = 0;
b8321b68	284	int number = rdev->raid_disk;
1da177e4 LT	285	struct multipath_info *p = conf->multipaths + number;
	286
	287	print_multipath_conf(conf);
	288
b8321b68	289	if (rdev == p->rdev) {
b2d444d7	290	if (test_bit(In_sync, &rdev->flags) \|\|
1da177e4	291	atomic_read(&rdev->nr_pending)) {
7279694d	292	pr_warn("hot-remove-disk, slot %d is identified but is still operational!\n", number);
1da177e4 LT	293	err = -EBUSY;
	294	goto abort;
	295	}
	296	p->rdev = NULL;
d787be40 N	297	if (!test_bit(RemoveSynchronized, &rdev->flags)) {
	298	synchronize_rcu();
	299	if (atomic_read(&rdev->nr_pending)) {
	300	/* lost the race, try later */
	301	err = -EBUSY;
	302	p->rdev = rdev;
	303	goto abort;
	304	}
1da177e4	305	}
a91a2785	306	err = md_integrity_register(mddev);
1da177e4 LT	307	}
	308	abort:
	309
	310	print_multipath_conf(conf);
	311	return err;
	312	}
	313
1da177e4 LT	314	/*
	315	* This is a kernel thread which:
	316	*
	317	* 1. Retries failed read operations on working multipaths.
	318	* 2. Updates the raid superblock when problems encounter.
	319	* 3. Performs writes following reads for array syncronising.
	320	*/
	321
4ed8731d	322	static void multipathd(struct md_thread *thread)
1da177e4	323	{
4ed8731d	324	struct mddev *mddev = thread->mddev;
1da177e4 LT	325	struct multipath_bh *mp_bh;
	326	struct bio *bio;
	327	unsigned long flags;
69724e28	328	struct mpconf *conf = mddev->private;
1da177e4 LT	329	struct list_head *head = &conf->retry_list;
	330
	331	md_check_recovery(mddev);
	332	for (;;) {
	333	char b[BDEVNAME_SIZE];
	334	spin_lock_irqsave(&conf->device_lock, flags);
	335	if (list_empty(head))
	336	break;
	337	mp_bh = list_entry(head->prev, struct multipath_bh, retry_list);
	338	list_del(head->prev);
	339	spin_unlock_irqrestore(&conf->device_lock, flags);
	340
	341	bio = &mp_bh->bio;
4f024f37	342	bio->bi_iter.bi_sector = mp_bh->master_bio->bi_iter.bi_sector;
f72ffdd6	343
1da177e4	344	if ((mp_bh->path = multipath_map (conf))<0) {
7279694d	345	pr_err("multipath: %s: unrecoverable IO read error for block %llu\n",
74d46992	346	bio_devname(bio, b),
7279694d	347	(unsigned long long)bio->bi_iter.bi_sector);
4e4cbee9	348	multipath_end_bh_io(mp_bh, BLK_STS_IOERR);
1da177e4	349	} else {
7279694d	350	pr_err("multipath: %s: redirecting sector %llu to another IO path\n",
74d46992	351	bio_devname(bio, b),
7279694d	352	(unsigned long long)bio->bi_iter.bi_sector);
1da177e4	353	bio = (mp_bh->master_bio);
4f024f37 KO	354	bio->bi_iter.bi_sector +=
4f024f37 KO	355	conf->multipaths[mp_bh->path].rdev->data_offset;
74d46992	356	bio_set_dev(bio, conf->multipaths[mp_bh->path].rdev->bdev);
1eff9d32	357	bio->bi_opf \|= REQ_FAILFAST_TRANSPORT;
1da177e4 LT	358	bio->bi_end_io = multipath_end_request;
	359	bio->bi_private = mp_bh;
	360	generic_make_request(bio);
	361	}
	362	}
	363	spin_unlock_irqrestore(&conf->device_lock, flags);
	364	}
	365
fd01b88c	366	static sector_t multipath_size(struct mddev *mddev, sector_t sectors, int raid_disks)
80c3a6ce DW	367	{
	368	WARN_ONCE(sectors \|\| raid_disks,
	369	"%s does not support generic reshape\n", __func__);
	370
	371	return mddev->dev_sectors;
	372	}
	373
fd01b88c	374	static int multipath_run (struct mddev *mddev)
1da177e4	375	{
69724e28	376	struct mpconf *conf;
1da177e4 LT	377	int disk_idx;
1da177e4 LT	378	struct multipath_info *disk;
3cb03002	379	struct md_rdev *rdev;
92f861a7	380	int working_disks;
afeee514	381	int ret;
1da177e4	382
0894cc30 AN	383	if (md_check_no_bitmap(mddev))
	384	return -EINVAL;
	385
1da177e4	386	if (mddev->level != LEVEL_MULTIPATH) {
7279694d N	387	pr_warn("multipath: %s: raid level not set to multipath IO (%d)\n",
7279694d N	388	mdname(mddev), mddev->level);
1da177e4 LT	389	goto out;
	390	}
	391	/*
	392	* copy the already verified devices into our private MULTIPATH
	393	* bookkeeping area. [whatever we allocate in multipath_run(),
afa0f557	394	* should be freed in multipath_free()]
1da177e4 LT	395	*/
1da177e4 LT	396
69724e28	397	conf = kzalloc(sizeof(struct mpconf), GFP_KERNEL);
1da177e4	398	mddev->private = conf;
7279694d	399	if (!conf)
1da177e4	400	goto out;
1da177e4	401
6396bb22 KC	402	conf->multipaths = kcalloc(mddev->raid_disks,
6396bb22 KC	403	sizeof(struct multipath_info),
1da177e4	404	GFP_KERNEL);
7279694d	405	if (!conf->multipaths)
1da177e4	406	goto out_free_conf;
1da177e4	407
92f861a7	408	working_disks = 0;
dafb20fa	409	rdev_for_each(rdev, mddev) {
1da177e4 LT	410	disk_idx = rdev->raid_disk;
	411	if (disk_idx < 0 \|\|
	412	disk_idx >= mddev->raid_disks)
	413	continue;
	414
	415	disk = conf->multipaths + disk_idx;
	416	disk->rdev = rdev;
8f6c2e4b MP	417	disk_stack_limits(mddev->gendisk, rdev->bdev,
8f6c2e4b MP	418	rdev->data_offset << 9);
1da177e4	419
b2d444d7	420	if (!test_bit(Faulty, &rdev->flags))
92f861a7	421	working_disks++;
1da177e4 LT	422	}
	423
	424	conf->raid_disks = mddev->raid_disks;
1da177e4 LT	425	conf->mddev = mddev;
	426	spin_lock_init(&conf->device_lock);
	427	INIT_LIST_HEAD(&conf->retry_list);
	428
92f861a7	429	if (!working_disks) {
7279694d	430	pr_warn("multipath: no operational IO paths for %s\n",
1da177e4 LT	431	mdname(mddev));
	432	goto out_free_conf;
	433	}
92f861a7	434	mddev->degraded = conf->raid_disks - working_disks;
1da177e4	435
afeee514 KO	436	ret = mempool_init_kmalloc_pool(&conf->pool, NR_RESERVED_BUFS,
	437	sizeof(struct multipath_bh));
	438	if (ret)
1da177e4	439	goto out_free_conf;
1da177e4	440
7279694d N	441	mddev->thread = md_register_thread(multipathd, mddev,
	442	"multipath");
	443	if (!mddev->thread)
	444	goto out_free_conf;
1da177e4	445
7279694d	446	pr_info("multipath: array %s active with %d out of %d IO paths\n",
92f861a7	447	mdname(mddev), conf->raid_disks - mddev->degraded,
7279694d	448	mddev->raid_disks);
1da177e4 LT	449	/*
	450	* Ok, everything is just fine now
	451	*/
1f403624	452	md_set_array_sectors(mddev, multipath_size(mddev, 0, 0));
7a5febe9	453
a91a2785 MP	454	if (md_integrity_register(mddev))
	455	goto out_free_conf;
	456
1da177e4 LT	457	return 0;
	458
	459	out_free_conf:
afeee514	460	mempool_exit(&conf->pool);
990a8baf	461	kfree(conf->multipaths);
1da177e4 LT	462	kfree(conf);
	463	mddev->private = NULL;
	464	out:
	465	return -EIO;
	466	}
	467
afa0f557	468	static void multipath_free(struct mddev mddev, void priv)
1da177e4	469	{
afa0f557	470	struct mpconf *conf = priv;
1da177e4	471
afeee514	472	mempool_exit(&conf->pool);
1da177e4 LT	473	kfree(conf->multipaths);
1da177e4 LT	474	kfree(conf);
1da177e4 LT	475	}
1da177e4 LT	476
84fc4b56	477	static struct md_personality multipath_personality =
1da177e4 LT	478	{
1da177e4 LT	479	.name = "multipath",
2604b703	480	.level = LEVEL_MULTIPATH,
1da177e4 LT	481	.owner = THIS_MODULE,
	482	.make_request = multipath_make_request,
	483	.run = multipath_run,
afa0f557	484	.free = multipath_free,
1da177e4 LT	485	.status = multipath_status,
	486	.error_handler = multipath_error,
	487	.hot_add_disk = multipath_add_disk,
	488	.hot_remove_disk= multipath_remove_disk,
80c3a6ce	489	.size = multipath_size,
5c675f83	490	.congested = multipath_congested,
1da177e4 LT	491	};
	492
	493	static int __init multipath_init (void)
	494	{
2604b703	495	return register_md_personality (&multipath_personality);
1da177e4 LT	496	}
	497
	498	static void __exit multipath_exit (void)
	499	{
2604b703	500	unregister_md_personality (&multipath_personality);
1da177e4 LT	501	}
	502
	503	module_init(multipath_init);
	504	module_exit(multipath_exit);
	505	MODULE_LICENSE("GPL");
0efb9e61	506	MODULE_DESCRIPTION("simple multi-path personality for MD");
1da177e4	507	MODULE_ALIAS("md-personality-7"); /* MULTIPATH */
d9d166c2	508	MODULE_ALIAS("md-multipath");
2604b703	509	MODULE_ALIAS("md-level--4");