[linux-2.6-block.git] / drivers / md / md-multipath.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * 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.
 */

#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
af1a8899	1	// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4 LT	2	/*
	3	* multipath.c : Multiple Devices driver for Linux
	4	*
	5	* Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
	6	*
	7	* Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
	8	*
	9	* MULTIPATH management functions.
	10	*
	11	* derived from raid1.c.
1da177e4 LT	12	*/
1da177e4 LT	13
bff61975	14	#include <linux/blkdev.h>
056075c7	15	#include <linux/module.h>
bff61975	16	#include <linux/raid/md_u.h>
bff61975	17	#include <linux/seq_file.h>
5a0e3ad6	18	#include <linux/slab.h>
43b2e5d8	19	#include "md.h"
935fe098	20	#include "md-multipath.h"
1da177e4 LT	21
	22	#define MAX_WORK_PER_DISK 128
	23
	24	#define NR_RESERVED_BUFS 32
	25
69724e28	26	static int multipath_map (struct mpconf *conf)
1da177e4 LT	27	{
	28	int i, disks = conf->raid_disks;
	29
	30	/*
f72ffdd6	31	* Later we do read balancing on the read side
1da177e4 LT	32	* now we use the first available disk.
	33	*/
	34
	35	rcu_read_lock();
	36	for (i = 0; i < disks; i++) {
3cb03002	37	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
f5b67ae8 N	38	if (rdev && test_bit(In_sync, &rdev->flags) &&
f5b67ae8 N	39	!test_bit(Faulty, &rdev->flags)) {
1da177e4 LT	40	atomic_inc(&rdev->nr_pending);
	41	rcu_read_unlock();
	42	return i;
	43	}
	44	}
	45	rcu_read_unlock();
	46
7279694d	47	pr_crit_ratelimited("multipath_map(): no more operational IO paths?\n");
1da177e4 LT	48	return (-1);
	49	}
	50
	51	static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
	52	{
	53	unsigned long flags;
fd01b88c	54	struct mddev *mddev = mp_bh->mddev;
69724e28	55	struct mpconf *conf = mddev->private;
1da177e4 LT	56
	57	spin_lock_irqsave(&conf->device_lock, flags);
	58	list_add(&mp_bh->retry_list, &conf->retry_list);
	59	spin_unlock_irqrestore(&conf->device_lock, flags);
	60	md_wakeup_thread(mddev->thread);
	61	}
	62
1da177e4 LT	63	/*
	64	* multipath_end_bh_io() is called when we have finished servicing a multipathed
	65	* operation and are ready to return a success/failure code to the buffer
	66	* cache layer.
	67	*/
4e4cbee9	68	static void multipath_end_bh_io(struct multipath_bh *mp_bh, blk_status_t status)
1da177e4 LT	69	{
1da177e4 LT	70	struct bio *bio = mp_bh->master_bio;
69724e28	71	struct mpconf *conf = mp_bh->mddev->private;
1da177e4	72
4e4cbee9	73	bio->bi_status = status;
4246a0b6	74	bio_endio(bio);
afeee514	75	mempool_free(mp_bh, &conf->pool);
1da177e4 LT	76	}
1da177e4 LT	77
4246a0b6	78	static void multipath_end_request(struct bio *bio)
1da177e4	79	{
7b92813c	80	struct multipath_bh *mp_bh = bio->bi_private;
69724e28	81	struct mpconf *conf = mp_bh->mddev->private;
3cb03002	82	struct md_rdev *rdev = conf->multipaths[mp_bh->path].rdev;
1da177e4	83
4e4cbee9	84	if (!bio->bi_status)
1da177e4	85	multipath_end_bh_io(mp_bh, 0);
1eff9d32	86	else if (!(bio->bi_opf & REQ_RAHEAD)) {
1da177e4 LT	87	/*
	88	* oops, IO error:
	89	*/
	90	char b[BDEVNAME_SIZE];
	91	md_error (mp_bh->mddev, rdev);
7279694d N	92	pr_info("multipath: %s: rescheduling sector %llu\n",
	93	bdevname(rdev->bdev,b),
	94	(unsigned long long)bio->bi_iter.bi_sector);
1da177e4 LT	95	multipath_reschedule_retry(mp_bh);
1da177e4 LT	96	} else
4e4cbee9	97	multipath_end_bh_io(mp_bh, bio->bi_status);
1da177e4	98	rdev_dec_pending(rdev, conf->mddev);
1da177e4 LT	99	}
1da177e4 LT	100
cc27b0c7	101	static bool multipath_make_request(struct mddev mddev, struct bio bio)
1da177e4	102	{
69724e28	103	struct mpconf *conf = mddev->private;
1da177e4 LT	104	struct multipath_bh * mp_bh;
	105	struct multipath_info *multipath;
	106
1eff9d32	107	if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
e9c7469b	108	md_flush_request(mddev, bio);
cc27b0c7	109	return true;
e5dcdd80 N	110	}
e5dcdd80 N	111
afeee514	112	mp_bh = mempool_alloc(&conf->pool, GFP_NOIO);
1da177e4 LT	113
	114	mp_bh->master_bio = bio;
	115	mp_bh->mddev = mddev;
	116
1da177e4 LT	117	mp_bh->path = multipath_map(conf);
1da177e4 LT	118	if (mp_bh->path < 0) {
4246a0b6	119	bio_io_error(bio);
afeee514	120	mempool_free(mp_bh, &conf->pool);
cc27b0c7	121	return true;
1da177e4 LT	122	}
	123	multipath = conf->multipaths + mp_bh->path;
	124
3a83f467	125	bio_init(&mp_bh->bio, NULL, 0);
fafcde3a ML	126	__bio_clone_fast(&mp_bh->bio, bio);
fafcde3a ML	127
4f024f37	128	mp_bh->bio.bi_iter.bi_sector += multipath->rdev->data_offset;
74d46992	129	bio_set_dev(&mp_bh->bio, multipath->rdev->bdev);
1eff9d32	130	mp_bh->bio.bi_opf \|= REQ_FAILFAST_TRANSPORT;
1da177e4 LT	131	mp_bh->bio.bi_end_io = multipath_end_request;
1da177e4 LT	132	mp_bh->bio.bi_private = mp_bh;
26483819	133	mddev_check_writesame(mddev, &mp_bh->bio);
3deff1a7	134	mddev_check_write_zeroes(mddev, &mp_bh->bio);
1da177e4	135	generic_make_request(&mp_bh->bio);
cc27b0c7	136	return true;
1da177e4 LT	137	}
1da177e4 LT	138
40cf2123	139	static void multipath_status(struct seq_file seq, struct mddev mddev)
1da177e4	140	{
69724e28	141	struct mpconf *conf = mddev->private;
1da177e4	142	int i;
f72ffdd6	143
1da177e4	144	seq_printf (seq, " [%d/%d] [", conf->raid_disks,
92f861a7	145	conf->raid_disks - mddev->degraded);
40cf2123 N	146	rcu_read_lock();
	147	for (i = 0; i < conf->raid_disks; i++) {
	148	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
	149	seq_printf (seq, "%s", rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	150	}
	151	rcu_read_unlock();
3acdb7b5	152	seq_putc(seq, ']');
1da177e4 LT	153	}
1da177e4 LT	154
5c675f83	155	static int multipath_congested(struct mddev *mddev, int bits)
0d129228	156	{
69724e28	157	struct mpconf *conf = mddev->private;
0d129228 N	158	int i, ret = 0;
	159
	160	rcu_read_lock();
	161	for (i = 0; i < mddev->raid_disks ; i++) {
3cb03002	162	struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
0d129228	163	if (rdev && !test_bit(Faulty, &rdev->flags)) {
165125e1	164	struct request_queue *q = bdev_get_queue(rdev->bdev);
0d129228	165
dc3b17cc	166	ret \|= bdi_congested(q->backing_dev_info, bits);
0d129228 N	167	/* Just like multipath_map, we just check the
	168	* first available device
	169	*/
	170	break;
	171	}
	172	}
	173	rcu_read_unlock();
	174	return ret;
	175	}
1da177e4 LT	176
	177	/*
	178	* Careful, this can execute in IRQ contexts as well!
	179	*/
fd01b88c	180	static void multipath_error (struct mddev mddev, struct md_rdev rdev)
1da177e4	181	{
69724e28	182	struct mpconf *conf = mddev->private;
6f8d0c77	183	char b[BDEVNAME_SIZE];
1da177e4	184
92f861a7	185	if (conf->raid_disks - mddev->degraded <= 1) {
1da177e4 LT	186	/*
	187	* Uh oh, we can do nothing if this is our last path, but
	188	* first check if this is a queued request for a device
	189	* which has just failed.
	190	*/
7279694d	191	pr_warn("multipath: only one IO path left and IO error.\n");
1da177e4	192	/* leave it active... it's all we have */
6f8d0c77 N	193	return;
	194	}
	195	/*
	196	* Mark disk as unusable
	197	*/
	198	if (test_and_clear_bit(In_sync, &rdev->flags)) {
	199	unsigned long flags;
	200	spin_lock_irqsave(&conf->device_lock, flags);
	201	mddev->degraded++;
	202	spin_unlock_irqrestore(&conf->device_lock, flags);
1da177e4	203	}
6f8d0c77	204	set_bit(Faulty, &rdev->flags);
2953079c	205	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
7279694d N	206	pr_err("multipath: IO failure on %s, disabling IO path.\n"
7279694d N	207	"multipath: Operation continuing on %d IO paths.\n",
6f8d0c77 N	208	bdevname(rdev->bdev, b),
6f8d0c77 N	209	conf->raid_disks - mddev->degraded);
1da177e4 LT	210	}
1da177e4 LT	211
69724e28	212	static void print_multipath_conf (struct mpconf *conf)
1da177e4 LT	213	{
	214	int i;
	215	struct multipath_info *tmp;
	216
7279694d	217	pr_debug("MULTIPATH conf printout:\n");
1da177e4	218	if (!conf) {
7279694d	219	pr_debug("(conf==NULL)\n");
1da177e4 LT	220	return;
1da177e4 LT	221	}
7279694d N	222	pr_debug(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
7279694d N	223	conf->raid_disks);
1da177e4 LT	224
	225	for (i = 0; i < conf->raid_disks; i++) {
	226	char b[BDEVNAME_SIZE];
	227	tmp = conf->multipaths + i;
	228	if (tmp->rdev)
7279694d N	229	pr_debug(" disk%d, o:%d, dev:%s\n",
	230	i,!test_bit(Faulty, &tmp->rdev->flags),
	231	bdevname(tmp->rdev->bdev,b));
1da177e4 LT	232	}
	233	}
	234
fd01b88c	235	static int multipath_add_disk(struct mddev mddev, struct md_rdev rdev)
1da177e4	236	{
69724e28	237	struct mpconf *conf = mddev->private;
199050ea	238	int err = -EEXIST;
1da177e4 LT	239	int path;
1da177e4 LT	240	struct multipath_info *p;
6c2fce2e NB	241	int first = 0;
	242	int last = mddev->raid_disks - 1;
	243
	244	if (rdev->raid_disk >= 0)
	245	first = last = rdev->raid_disk;
1da177e4 LT	246
	247	print_multipath_conf(conf);
	248
6c2fce2e	249	for (path = first; path <= last; path++)
1da177e4	250	if ((p=conf->multipaths+path)->rdev == NULL) {
8f6c2e4b MP	251	disk_stack_limits(mddev->gendisk, rdev->bdev,
8f6c2e4b MP	252	rdev->data_offset << 9);
1da177e4	253
1501efad DW	254	err = md_integrity_add_rdev(rdev, mddev);
	255	if (err)
	256	break;
6f8d0c77	257	spin_lock_irq(&conf->device_lock);
750a8f3e	258	mddev->degraded--;
1da177e4	259	rdev->raid_disk = path;
b2d444d7	260	set_bit(In_sync, &rdev->flags);
6f8d0c77	261	spin_unlock_irq(&conf->device_lock);
d6065f7b	262	rcu_assign_pointer(p->rdev, rdev);
199050ea NB	263	err = 0;
199050ea NB	264	break;
1da177e4 LT	265	}
	266
	267	print_multipath_conf(conf);
199050ea NB	268
199050ea NB	269	return err;
1da177e4 LT	270	}
1da177e4 LT	271
b8321b68	272	static int multipath_remove_disk(struct mddev mddev, struct md_rdev rdev)
1da177e4	273	{
69724e28	274	struct mpconf *conf = mddev->private;
1da177e4	275	int err = 0;
b8321b68	276	int number = rdev->raid_disk;
1da177e4 LT	277	struct multipath_info *p = conf->multipaths + number;
	278
	279	print_multipath_conf(conf);
	280
b8321b68	281	if (rdev == p->rdev) {
b2d444d7	282	if (test_bit(In_sync, &rdev->flags) \|\|
1da177e4	283	atomic_read(&rdev->nr_pending)) {
7279694d	284	pr_warn("hot-remove-disk, slot %d is identified but is still operational!\n", number);
1da177e4 LT	285	err = -EBUSY;
	286	goto abort;
	287	}
	288	p->rdev = NULL;
d787be40 N	289	if (!test_bit(RemoveSynchronized, &rdev->flags)) {
	290	synchronize_rcu();
	291	if (atomic_read(&rdev->nr_pending)) {
	292	/* lost the race, try later */
	293	err = -EBUSY;
	294	p->rdev = rdev;
	295	goto abort;
	296	}
1da177e4	297	}
a91a2785	298	err = md_integrity_register(mddev);
1da177e4 LT	299	}
	300	abort:
	301
	302	print_multipath_conf(conf);
	303	return err;
	304	}
	305
1da177e4 LT	306	/*
	307	* This is a kernel thread which:
	308	*
	309	* 1. Retries failed read operations on working multipaths.
	310	* 2. Updates the raid superblock when problems encounter.
	311	* 3. Performs writes following reads for array syncronising.
	312	*/
	313
4ed8731d	314	static void multipathd(struct md_thread *thread)
1da177e4	315	{
4ed8731d	316	struct mddev *mddev = thread->mddev;
1da177e4 LT	317	struct multipath_bh *mp_bh;
	318	struct bio *bio;
	319	unsigned long flags;
69724e28	320	struct mpconf *conf = mddev->private;
1da177e4 LT	321	struct list_head *head = &conf->retry_list;
	322
	323	md_check_recovery(mddev);
	324	for (;;) {
	325	char b[BDEVNAME_SIZE];
	326	spin_lock_irqsave(&conf->device_lock, flags);
	327	if (list_empty(head))
	328	break;
	329	mp_bh = list_entry(head->prev, struct multipath_bh, retry_list);
	330	list_del(head->prev);
	331	spin_unlock_irqrestore(&conf->device_lock, flags);
	332
	333	bio = &mp_bh->bio;
4f024f37	334	bio->bi_iter.bi_sector = mp_bh->master_bio->bi_iter.bi_sector;
f72ffdd6	335
1da177e4	336	if ((mp_bh->path = multipath_map (conf))<0) {
7279694d	337	pr_err("multipath: %s: unrecoverable IO read error for block %llu\n",
74d46992	338	bio_devname(bio, b),
7279694d	339	(unsigned long long)bio->bi_iter.bi_sector);
4e4cbee9	340	multipath_end_bh_io(mp_bh, BLK_STS_IOERR);
1da177e4	341	} else {
7279694d	342	pr_err("multipath: %s: redirecting sector %llu to another IO path\n",
74d46992	343	bio_devname(bio, b),
7279694d	344	(unsigned long long)bio->bi_iter.bi_sector);
1da177e4	345	bio = (mp_bh->master_bio);
4f024f37 KO	346	bio->bi_iter.bi_sector +=
4f024f37 KO	347	conf->multipaths[mp_bh->path].rdev->data_offset;
74d46992	348	bio_set_dev(bio, conf->multipaths[mp_bh->path].rdev->bdev);
1eff9d32	349	bio->bi_opf \|= REQ_FAILFAST_TRANSPORT;
1da177e4 LT	350	bio->bi_end_io = multipath_end_request;
	351	bio->bi_private = mp_bh;
	352	generic_make_request(bio);
	353	}
	354	}
	355	spin_unlock_irqrestore(&conf->device_lock, flags);
	356	}
	357
fd01b88c	358	static sector_t multipath_size(struct mddev *mddev, sector_t sectors, int raid_disks)
80c3a6ce DW	359	{
	360	WARN_ONCE(sectors \|\| raid_disks,
	361	"%s does not support generic reshape\n", __func__);
	362
	363	return mddev->dev_sectors;
	364	}
	365
fd01b88c	366	static int multipath_run (struct mddev *mddev)
1da177e4	367	{
69724e28	368	struct mpconf *conf;
1da177e4 LT	369	int disk_idx;
1da177e4 LT	370	struct multipath_info *disk;
3cb03002	371	struct md_rdev *rdev;
92f861a7	372	int working_disks;
afeee514	373	int ret;
1da177e4	374
0894cc30 AN	375	if (md_check_no_bitmap(mddev))
	376	return -EINVAL;
	377
1da177e4	378	if (mddev->level != LEVEL_MULTIPATH) {
7279694d N	379	pr_warn("multipath: %s: raid level not set to multipath IO (%d)\n",
7279694d N	380	mdname(mddev), mddev->level);
1da177e4 LT	381	goto out;
	382	}
	383	/*
	384	* copy the already verified devices into our private MULTIPATH
	385	* bookkeeping area. [whatever we allocate in multipath_run(),
afa0f557	386	* should be freed in multipath_free()]
1da177e4 LT	387	*/
1da177e4 LT	388
69724e28	389	conf = kzalloc(sizeof(struct mpconf), GFP_KERNEL);
1da177e4	390	mddev->private = conf;
7279694d	391	if (!conf)
1da177e4	392	goto out;
1da177e4	393
6396bb22 KC	394	conf->multipaths = kcalloc(mddev->raid_disks,
6396bb22 KC	395	sizeof(struct multipath_info),
1da177e4	396	GFP_KERNEL);
7279694d	397	if (!conf->multipaths)
1da177e4	398	goto out_free_conf;
1da177e4	399
92f861a7	400	working_disks = 0;
dafb20fa	401	rdev_for_each(rdev, mddev) {
1da177e4 LT	402	disk_idx = rdev->raid_disk;
	403	if (disk_idx < 0 \|\|
	404	disk_idx >= mddev->raid_disks)
	405	continue;
	406
	407	disk = conf->multipaths + disk_idx;
	408	disk->rdev = rdev;
8f6c2e4b MP	409	disk_stack_limits(mddev->gendisk, rdev->bdev,
8f6c2e4b MP	410	rdev->data_offset << 9);
1da177e4	411
b2d444d7	412	if (!test_bit(Faulty, &rdev->flags))
92f861a7	413	working_disks++;
1da177e4 LT	414	}
	415
	416	conf->raid_disks = mddev->raid_disks;
1da177e4 LT	417	conf->mddev = mddev;
	418	spin_lock_init(&conf->device_lock);
	419	INIT_LIST_HEAD(&conf->retry_list);
	420
92f861a7	421	if (!working_disks) {
7279694d	422	pr_warn("multipath: no operational IO paths for %s\n",
1da177e4 LT	423	mdname(mddev));
	424	goto out_free_conf;
	425	}
92f861a7	426	mddev->degraded = conf->raid_disks - working_disks;
1da177e4	427
afeee514 KO	428	ret = mempool_init_kmalloc_pool(&conf->pool, NR_RESERVED_BUFS,
	429	sizeof(struct multipath_bh));
	430	if (ret)
1da177e4	431	goto out_free_conf;
1da177e4	432
7279694d N	433	mddev->thread = md_register_thread(multipathd, mddev,
	434	"multipath");
	435	if (!mddev->thread)
	436	goto out_free_conf;
1da177e4	437
7279694d	438	pr_info("multipath: array %s active with %d out of %d IO paths\n",
92f861a7	439	mdname(mddev), conf->raid_disks - mddev->degraded,
7279694d	440	mddev->raid_disks);
1da177e4 LT	441	/*
	442	* Ok, everything is just fine now
	443	*/
1f403624	444	md_set_array_sectors(mddev, multipath_size(mddev, 0, 0));
7a5febe9	445
a91a2785 MP	446	if (md_integrity_register(mddev))
	447	goto out_free_conf;
	448
1da177e4 LT	449	return 0;
	450
	451	out_free_conf:
afeee514	452	mempool_exit(&conf->pool);
990a8baf	453	kfree(conf->multipaths);
1da177e4 LT	454	kfree(conf);
	455	mddev->private = NULL;
	456	out:
	457	return -EIO;
	458	}
	459
afa0f557	460	static void multipath_free(struct mddev mddev, void priv)
1da177e4	461	{
afa0f557	462	struct mpconf *conf = priv;
1da177e4	463
afeee514	464	mempool_exit(&conf->pool);
1da177e4 LT	465	kfree(conf->multipaths);
1da177e4 LT	466	kfree(conf);
1da177e4 LT	467	}
1da177e4 LT	468
84fc4b56	469	static struct md_personality multipath_personality =
1da177e4 LT	470	{
1da177e4 LT	471	.name = "multipath",
2604b703	472	.level = LEVEL_MULTIPATH,
1da177e4 LT	473	.owner = THIS_MODULE,
	474	.make_request = multipath_make_request,
	475	.run = multipath_run,
afa0f557	476	.free = multipath_free,
1da177e4 LT	477	.status = multipath_status,
	478	.error_handler = multipath_error,
	479	.hot_add_disk = multipath_add_disk,
	480	.hot_remove_disk= multipath_remove_disk,
80c3a6ce	481	.size = multipath_size,
5c675f83	482	.congested = multipath_congested,
1da177e4 LT	483	};
	484
	485	static int __init multipath_init (void)
	486	{
2604b703	487	return register_md_personality (&multipath_personality);
1da177e4 LT	488	}
	489
	490	static void __exit multipath_exit (void)
	491	{
2604b703	492	unregister_md_personality (&multipath_personality);
1da177e4 LT	493	}
	494
	495	module_init(multipath_init);
	496	module_exit(multipath_exit);
	497	MODULE_LICENSE("GPL");
0efb9e61	498	MODULE_DESCRIPTION("simple multi-path personality for MD");
1da177e4	499	MODULE_ALIAS("md-personality-7"); /* MULTIPATH */
d9d166c2	500	MODULE_ALIAS("md-multipath");
2604b703	501	MODULE_ALIAS("md-level--4");