multipath.c

linux和2410结合开发 用他可以生成2410所需的zImage文件

/*
 * 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/raid/multipath.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


/*
 * The following can be used to debug the driver
 */
#define MULTIPATH_DEBUG	0

#if MULTIPATH_DEBUG
#define PRINTK(x...)   printk(x)
#define inline
#define __inline__
#else
#define PRINTK(x...)  do { } while (0)
#endif


static mdk_personality_t multipath_personality;
static md_spinlock_t retry_list_lock = MD_SPIN_LOCK_UNLOCKED;
struct multipath_bh *multipath_retry_list = NULL, **multipath_retry_tail;

static int multipath_diskop(mddev_t *mddev, mdp_disk_t **d, int state);



static struct multipath_bh *multipath_alloc_mpbh(multipath_conf_t *conf)
{
	struct multipath_bh *mp_bh = NULL;

	do {
		md_spin_lock_irq(&conf->device_lock);
		if (!conf->freer1_blocked && conf->freer1) {
			mp_bh = conf->freer1;
			conf->freer1 = mp_bh->next_mp;
			conf->freer1_cnt--;
			mp_bh->next_mp = NULL;
			mp_bh->state = (1 << MPBH_PreAlloc);
			mp_bh->bh_req.b_state = 0;
		}
		md_spin_unlock_irq(&conf->device_lock);
		if (mp_bh)
			return mp_bh;
		mp_bh = (struct multipath_bh *) kmalloc(sizeof(struct multipath_bh),
					GFP_NOIO);
		if (mp_bh) {
			memset(mp_bh, 0, sizeof(*mp_bh));
			return mp_bh;
		}
		conf->freer1_blocked = 1;
		wait_disk_event(conf->wait_buffer,
				!conf->freer1_blocked ||
				conf->freer1_cnt > NR_RESERVED_BUFS/2
		    );
		conf->freer1_blocked = 0;
	} while (1);
}

static inline void multipath_free_mpbh(struct multipath_bh *mp_bh)
{
	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);

	if (test_bit(MPBH_PreAlloc, &mp_bh->state)) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		mp_bh->next_mp = conf->freer1;
		conf->freer1 = mp_bh;
		conf->freer1_cnt++;
		spin_unlock_irqrestore(&conf->device_lock, flags);
		wake_up(&conf->wait_buffer);
	} else {
		kfree(mp_bh);
	}
}

static int multipath_grow_mpbh (multipath_conf_t *conf, int cnt)
{
	int i = 0;

	while (i < cnt) {
		struct multipath_bh *mp_bh;
		mp_bh = (struct multipath_bh*)kmalloc(sizeof(*mp_bh), GFP_KERNEL);
		if (!mp_bh)
			break;
		memset(mp_bh, 0, sizeof(*mp_bh));
		set_bit(MPBH_PreAlloc, &mp_bh->state);
		mp_bh->mddev = conf->mddev;	       

		multipath_free_mpbh(mp_bh);
		i++;
	}
	return i;
}

static void multipath_shrink_mpbh(multipath_conf_t *conf)
{
	md_spin_lock_irq(&conf->device_lock);
	while (conf->freer1) {
		struct multipath_bh *mp_bh = conf->freer1;
		conf->freer1 = mp_bh->next_mp;
		conf->freer1_cnt--;
		kfree(mp_bh);
	}
	md_spin_unlock_irq(&conf->device_lock);
}


static int multipath_map (mddev_t *mddev, kdev_t *rdev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	int i, disks = MD_SB_DISKS;

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

	for (i = 0; i < disks; i++) {
		if (conf->multipaths[i].operational) {
			*rdev = conf->multipaths[i].dev;
			return (0);
		}
	}

	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);

	md_spin_lock_irqsave(&retry_list_lock, flags);
	if (multipath_retry_list == NULL)
		multipath_retry_tail = &multipath_retry_list;
	*multipath_retry_tail = mp_bh;
	multipath_retry_tail = &mp_bh->next_mp;
	mp_bh->next_mp = NULL;
	md_spin_unlock_irqrestore(&retry_list_lock, flags);
	md_wakeup_thread(conf->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 uptodate)
{
	struct buffer_head *bh = mp_bh->master_bh;

	bh->b_end_io(bh, uptodate);
	multipath_free_mpbh(mp_bh);
}

void multipath_end_request (struct buffer_head *bh, int uptodate)
{
	struct multipath_bh * mp_bh = (struct multipath_bh *)(bh->b_private);

	/*
	 * this branch is our 'one multipath IO has finished' event handler:
	 */
	if (!uptodate)
		md_error (mp_bh->mddev, bh->b_dev);
	else
		/*
		 * Set MPBH_Uptodate in our master buffer_head, so that
		 * we will return a good error code for to the higher
		 * levels even if IO on some other multipathed buffer fails.
		 *
		 * The 'master' represents the complex operation to 
		 * user-side. So if something waits for IO, then it will
		 * wait for the 'master' buffer_head.
		 */
		set_bit (MPBH_Uptodate, &mp_bh->state);

		
	if (uptodate) {
		multipath_end_bh_io(mp_bh, uptodate);
		return;
	}
	/*
	 * oops, IO error:
	 */
	printk(KERN_ERR "multipath: %s: rescheduling block %lu\n", 
		 partition_name(bh->b_dev), bh->b_blocknr);
	multipath_reschedule_retry(mp_bh);
	return;
}

/*
 * This routine returns the disk from which the requested read should
 * be done.
 */

static int multipath_read_balance (multipath_conf_t *conf)
{
	int disk;

	for (disk = 0; disk < conf->raid_disks; disk++)	
		if (conf->multipaths[disk].operational)
			return disk;
	BUG();
	return 0;
}

static int multipath_make_request (mddev_t *mddev, int rw,
			       struct buffer_head * bh)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	struct buffer_head *bh_req;
	struct multipath_bh * mp_bh;
	struct multipath_info *multipath;

	if (!buffer_locked(bh))
		BUG();
	
/*
 * make_request() can abort the operation when READA is being
 * used and no empty request is available.
 *
 * Currently, just replace the command with READ/WRITE.
 */
	if (rw == READA)
		rw = READ;

	mp_bh = multipath_alloc_mpbh (conf);

	mp_bh->master_bh = bh;
	mp_bh->mddev = mddev;
	mp_bh->cmd = rw;

	/*
	 * read balancing logic:
	 */
	multipath = conf->multipaths + multipath_read_balance(conf);

	bh_req = &mp_bh->bh_req;
	memcpy(bh_req, bh, sizeof(*bh));
	bh_req->b_blocknr = bh->b_rsector;
	bh_req->b_dev = multipath->dev;
	bh_req->b_rdev = multipath->dev;
/*	bh_req->b_rsector = bh->n_rsector; */
	bh_req->b_end_io = multipath_end_request;
	bh_req->b_private = mp_bh;
	generic_make_request (rw, bh_req);
	return 0;
}

static int multipath_status (char *page, mddev_t *mddev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	int sz = 0, i;
	
	sz += sprintf (page+sz, " [%d/%d] [", conf->raid_disks,
						 conf->working_disks);
	for (i = 0; i < conf->raid_disks; i++)
		sz += sprintf (page+sz, "%s",
			conf->multipaths[i].operational ? "U" : "_");
	sz += sprintf (page+sz, "]");
	return sz;
}

#define LAST_DISK KERN_ALERT \
"multipath: only one IO path left and IO error.\n"

#define NO_SPARE_DISK KERN_ALERT \
"multipath: no spare IO path left!\n"

#define DISK_FAILED KERN_ALERT \
"multipath: IO failure on %s, disabling IO path. \n" \
"	Operation continuing on %d IO paths.\n"

static void mark_disk_bad (mddev_t *mddev, int failed)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	struct multipath_info *multipath = conf->multipaths+failed;
	mdp_super_t *sb = mddev->sb;

	multipath->operational = 0;
	mark_disk_faulty(sb->disks+multipath->number);
	mark_disk_nonsync(sb->disks+multipath->number);
	mark_disk_inactive(sb->disks+multipath->number);
	sb->active_disks--;
	sb->working_disks--;
	sb->failed_disks++;
	mddev->sb_dirty = 1;
	md_wakeup_thread(conf->thread);
	conf->working_disks--;
	printk (DISK_FAILED, partition_name (multipath->dev),
				 conf->working_disks);
}

/*
 * Careful, this can execute in IRQ contexts as well!
 */
static int multipath_error (mddev_t *mddev, kdev_t dev)
{
	multipath_conf_t *conf = mddev_to_conf(mddev);
	struct multipath_info * multipaths = conf->multipaths;
	int disks = MD_SB_DISKS;
	int other_paths = 1;
	int i;

	if (conf->working_disks == 1) {
		other_paths = 0;
		for (i = 0; i < disks; i++) {
			if (multipaths[i].spare) {
				other_paths = 1;
				break;
			}
		}
	}

	if (!other_paths) {
		/*
		 * 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.
		 */
		for (i = 0; i < disks; i++) {
			if (multipaths[i].dev==dev && !multipaths[i].operational)
				return 0;
		}
		printk (LAST_DISK);
	} else {
		/*
		 * Mark disk as unusable
		 */
		for (i = 0; i < disks; i++) {
			if (multipaths[i].dev==dev && multipaths[i].operational) {
				mark_disk_bad(mddev, i);
				break;
			}
		}
		if (!conf->working_disks) {
			int err = 1;
			mdp_disk_t *spare;
			mdp_super_t *sb = mddev->sb;

			spare = get_spare(mddev);
			if (spare) {
				err = multipath_diskop(mddev, &spare, DISKOP_SPARE_WRITE);
				printk("got DISKOP_SPARE_WRITE err: %d. (spare_faulty(): %d)\n", err, disk_faulty(spare));
			}
			if (!err && !disk_faulty(spare)) {
				multipath_diskop(mddev, &spare, DISKOP_SPARE_ACTIVE);
				mark_disk_sync(spare);
				mark_disk_active(spare);
				sb->active_disks++;
				sb->spare_disks--;
			}
		}
	}
	return 0;
}

#undef LAST_DISK
#undef NO_SPARE_DISK
#undef DISK_FAILED


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 nd:%d\n", conf->working_disks,
			 conf->raid_disks, conf->nr_disks);

	for (i = 0; i < MD_SB_DISKS; i++) {
		tmp = conf->multipaths + i;
		if (tmp->spare || tmp->operational || tmp->number ||
				tmp->raid_disk || tmp->used_slot)
			printk(" disk%d, s:%d, o:%d, n:%d rd:%d us:%d dev:%s\n",
				i, tmp->spare,tmp->operational,
				tmp->number,tmp->raid_disk,tmp->used_slot,
				partition_name(tmp->dev));
	}
}

static int multipath_diskop(mddev_t *mddev, mdp_disk_t **d, int state)
{
	int err = 0;
	int i, failed_disk=-1, spare_disk=-1, removed_disk=-1, added_disk=-1;
	multipath_conf_t *conf = mddev->private;
	struct multipath_info *tmp, *sdisk, *fdisk, *rdisk, *adisk;
	mdp_super_t *sb = mddev->sb;
	mdp_disk_t *failed_desc, *spare_desc, *added_desc;
	mdk_rdev_t *spare_rdev, *failed_rdev;

	print_multipath_conf(conf);
	md_spin_lock_irq(&conf->device_lock);
	/*
	 * find the disk ...
	 */
	switch (state) {

	case DISKOP_SPARE_ACTIVE:

		/*
		 * Find the failed disk within the MULTIPATH configuration ...
		 * (this can only be in the first conf->working_disks part)
		 */
		for (i = 0; i < conf->raid_disks; i++) {
			tmp = conf->multipaths + i;
			if ((!tmp->operational && !tmp->spare) ||
					!tmp->used_slot) {
				failed_disk = i;
				break;
			}
		}
		/*
		 * When we activate a spare disk we _must_ have a disk in
		 * the lower (active) part of the array to replace. 
		 */
		if ((failed_disk == -1) || (failed_disk >= conf->raid_disks)) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		/* fall through */

	case DISKOP_SPARE_WRITE:
	case DISKOP_SPARE_INACTIVE:

		/*
		 * Find the spare disk ... (can only be in the 'high'
		 * area of the array)
		 */
		for (i = conf->raid_disks; i < MD_SB_DISKS; i++) {
			tmp = conf->multipaths + i;
			if (tmp->spare && tmp->number == (*d)->number) {
				spare_disk = i;
				break;
			}
		}
		if (spare_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;

	case DISKOP_HOT_REMOVE_DISK:

		for (i = 0; i < MD_SB_DISKS; i++) {
			tmp = conf->multipaths + i;
			if (tmp->used_slot && (tmp->number == (*d)->number)) {
				if (tmp->operational) {
					printk(KERN_ERR "hot-remove-disk, slot %d is identified to be the requested disk (number %d), but is still operational!\n", i, (*d)->number);
					err = -EBUSY;
					goto abort;
				}
				removed_disk = i;
				break;
			}
		}
		if (removed_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;

	case DISKOP_HOT_ADD_DISK:

		for (i = conf->raid_disks; i < MD_SB_DISKS; i++) {
			tmp = conf->multipaths + i;
			if (!tmp->used_slot) {
				added_disk = i;
				break;
			}
		}
		if (added_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;
	}

	switch (state) {
	/*
	 * Switch the spare disk to write-only mode:
	 */
	case DISKOP_SPARE_WRITE:
		sdisk = conf->multipaths + spare_disk;
		sdisk->operational = 1;
		break;
	/*
	 * Deactivate a spare disk:
	 */
	case DISKOP_SPARE_INACTIVE:
		sdisk = conf->multipaths + spare_disk;
		sdisk->operational = 0;
		break;