patch-2.4.10 linux/drivers/md/multipath.c

Next file: linux/drivers/md/raid1.c
Previous file: linux/drivers/md/md.c
Back to the patch index
Back to the overall index

diff -u --recursive --new-file v2.4.9/linux/drivers/md/multipath.c linux/drivers/md/multipath.c
@@ -0,0 +1,1261 @@
+/*
+ * 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.
+ *
+ * Better read-balancing code written by Mika Kuoppala <miku@iki.fi>, 2000
+ *
+ * Fixes to reconstruction by Jakob Østergaard" <jakob@ostenfeld.dk>
+ * Various fixes by Neil Brown <neilb@cse.unsw.edu.au>
+ *
+ * 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
+
+/*
+ * 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);
+
+struct buffer_head *multipath_alloc_bh(multipath_conf_t *conf, int cnt)
+{
+	/* return a linked list of "cnt" struct buffer_heads.
+	 * don't take any off the free list unless we know we can
+	 * get all we need, otherwise we could deadlock
+	 */
+	struct buffer_head *bh=NULL;
+
+	while(cnt) {
+		struct buffer_head *t;
+		md_spin_lock_irq(&conf->device_lock);
+		if (conf->freebh_cnt >= cnt)
+			while (cnt) {
+				t = conf->freebh;
+				conf->freebh = t->b_next;
+				t->b_next = bh;
+				bh = t;
+				t->b_state = 0;
+				conf->freebh_cnt--;
+				cnt--;
+			}
+		md_spin_unlock_irq(&conf->device_lock);
+		if (cnt == 0)
+			break;
+		t = (struct buffer_head *)kmalloc(sizeof(struct buffer_head), GFP_NOIO);
+		if (t) {
+			memset(t, 0, sizeof(*t));
+			t->b_next = bh;
+			bh = t;
+			cnt--;
+		} else {
+			PRINTK("waiting for %d bh\n", cnt);
+			wait_event(conf->wait_buffer, conf->freebh_cnt >= cnt);
+		}
+	}
+	return bh;
+}
+
+static inline void multipath_free_bh(multipath_conf_t *conf, struct buffer_head *bh)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&conf->device_lock, flags);
+	while (bh) {
+		struct buffer_head *t = bh;
+		bh=bh->b_next;
+		if (t->b_pprev == NULL)
+			kfree(t);
+		else {
+			t->b_next= conf->freebh;
+			conf->freebh = t;
+			conf->freebh_cnt++;
+		}
+	}
+	spin_unlock_irqrestore(&conf->device_lock, flags);
+	wake_up(&conf->wait_buffer);
+}
+
+static int multipath_grow_bh(multipath_conf_t *conf, int cnt)
+{
+	/* allocate cnt buffer_heads, possibly less if kalloc fails */
+	int i = 0;
+
+	while (i < cnt) {
+		struct buffer_head *bh;
+		bh = kmalloc(sizeof(*bh), GFP_KERNEL);
+		if (!bh) break;
+		memset(bh, 0, sizeof(*bh));
+
+		md_spin_lock_irq(&conf->device_lock);
+		bh->b_pprev = &conf->freebh;
+		bh->b_next = conf->freebh;
+		conf->freebh = bh;
+		conf->freebh_cnt++;
+		md_spin_unlock_irq(&conf->device_lock);
+
+		i++;
+	}
+	return i;
+}
+
+static int multipath_shrink_bh(multipath_conf_t *conf, int cnt)
+{
+	/* discard cnt buffer_heads, if we can find them */
+	int i = 0;
+
+	md_spin_lock_irq(&conf->device_lock);
+	while ((i < cnt) && conf->freebh) {
+		struct buffer_head *bh = conf->freebh;
+		conf->freebh = bh->b_next;
+		kfree(bh);
+		i++;
+		conf->freebh_cnt--;
+	}
+	md_spin_unlock_irq(&conf->device_lock);
+	return i;
+}
+		
+
+static struct multipath_bh *multipath_alloc_mpbh(multipath_conf_t *conf)
+{
+	struct multipath_bh *r1_bh = NULL;
+
+	do {
+		md_spin_lock_irq(&conf->device_lock);
+		if (conf->freer1) {
+			r1_bh = conf->freer1;
+			conf->freer1 = r1_bh->next_r1;
+			r1_bh->next_r1 = NULL;
+			r1_bh->state = 0;
+			r1_bh->bh_req.b_state = 0;
+		}
+		md_spin_unlock_irq(&conf->device_lock);
+		if (r1_bh)
+			return r1_bh;
+		r1_bh = (struct multipath_bh *) kmalloc(sizeof(struct multipath_bh),
+					GFP_NOIO);
+		if (r1_bh) {
+			memset(r1_bh, 0, sizeof(*r1_bh));
+			return r1_bh;
+		}
+		wait_event(conf->wait_buffer, conf->freer1);
+	} while (1);
+}
+
+static inline void multipath_free_mpbh(struct multipath_bh *r1_bh)
+{
+	struct buffer_head *bh = r1_bh->multipath_bh_list;
+	multipath_conf_t *conf = mddev_to_conf(r1_bh->mddev);
+
+	r1_bh->multipath_bh_list = NULL;
+
+	if (test_bit(MPBH_PreAlloc, &r1_bh->state)) {
+		unsigned long flags;
+		spin_lock_irqsave(&conf->device_lock, flags);
+		r1_bh->next_r1 = conf->freer1;
+		conf->freer1 = r1_bh;
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+	} else {
+		kfree(r1_bh);
+	}
+	multipath_free_bh(conf, bh);
+}
+
+static int multipath_grow_mpbh (multipath_conf_t *conf, int cnt)
+{
+	int i = 0;
+
+	while (i < cnt) {
+		struct multipath_bh *r1_bh;
+		r1_bh = (struct multipath_bh*)kmalloc(sizeof(*r1_bh), GFP_KERNEL);
+		if (!r1_bh)
+			break;
+		memset(r1_bh, 0, sizeof(*r1_bh));
+
+		md_spin_lock_irq(&conf->device_lock);
+		set_bit(MPBH_PreAlloc, &r1_bh->state);
+		r1_bh->next_r1 = conf->freer1;
+		conf->freer1 = r1_bh;
+		md_spin_unlock_irq(&conf->device_lock);
+
+		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 *r1_bh = conf->freer1;
+		conf->freer1 = r1_bh->next_r1;
+		kfree(r1_bh);
+	}
+	md_spin_unlock_irq(&conf->device_lock);
+}
+
+
+
+static inline void multipath_free_buf(struct multipath_bh *r1_bh)
+{
+	unsigned long flags;
+	struct buffer_head *bh = r1_bh->multipath_bh_list;
+	multipath_conf_t *conf = mddev_to_conf(r1_bh->mddev);
+	r1_bh->multipath_bh_list = NULL;
+	
+	spin_lock_irqsave(&conf->device_lock, flags);
+	r1_bh->next_r1 = conf->freebuf;
+	conf->freebuf = r1_bh;
+	spin_unlock_irqrestore(&conf->device_lock, flags);
+	multipath_free_bh(conf, bh);
+}
+
+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 *r1_bh)
+{
+	unsigned long flags;
+	mddev_t *mddev = r1_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 = r1_bh;
+	multipath_retry_tail = &r1_bh->next_r1;
+	r1_bh->next_r1 = NULL;
+	md_spin_unlock_irqrestore(&retry_list_lock, flags);
+	md_wakeup_thread(conf->thread);
+}
+
+
+static void inline io_request_done(unsigned long sector, multipath_conf_t *conf, int phase)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&conf->segment_lock, flags);
+	if (sector < conf->start_active)
+		conf->cnt_done--;
+	else if (sector >= conf->start_future && conf->phase == phase)
+		conf->cnt_future--;
+	else if (!--conf->cnt_pending)
+		wake_up(&conf->wait_ready);
+
+	spin_unlock_irqrestore(&conf->segment_lock, flags);
+}
+
+static void inline sync_request_done (unsigned long sector, multipath_conf_t *conf)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&conf->segment_lock, flags);
+	if (sector >= conf->start_ready)
+		--conf->cnt_ready;
+	else if (sector >= conf->start_active) {
+		if (!--conf->cnt_active) {
+			conf->start_active = conf->start_ready;
+			wake_up(&conf->wait_done);
+		}
+	}
+	spin_unlock_irqrestore(&conf->segment_lock, flags);
+}
+
+/*
+ * 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 *r1_bh, int uptodate)
+{
+	struct buffer_head *bh = r1_bh->master_bh;
+
+	io_request_done(bh->b_rsector, mddev_to_conf(r1_bh->mddev),
+			test_bit(MPBH_SyncPhase, &r1_bh->state));
+
+	bh->b_end_io(bh, uptodate);
+	multipath_free_mpbh(r1_bh);
+}
+
+void multipath_end_request (struct buffer_head *bh, int uptodate)
+{
+	struct multipath_bh * r1_bh = (struct multipath_bh *)(bh->b_private);
+
+	/*
+	 * this branch is our 'one multipath IO has finished' event handler:
+	 */
+	if (!uptodate)
+		md_error (r1_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, &r1_bh->state);
+
+		
+	if (uptodate) {
+		multipath_end_bh_io(r1_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(r1_bh);
+	return;
+}
+
+/*
+ * This routine returns the disk from which the requested read should
+ * be done. It bookkeeps the last read position for every disk
+ * in array and when new read requests come, the disk which last
+ * position is nearest to the request, is chosen.
+ *
+ * TODO: now if there are 2 multipaths in the same 2 devices, performance
+ * degrades dramatically because position is multipath, not device based.
+ * This should be changed to be device based. Also atomic sequential
+ * reads should be somehow balanced.
+ */
+
+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 * r1_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;
+
+	r1_bh = multipath_alloc_mpbh (conf);
+
+	spin_lock_irq(&conf->segment_lock);
+	wait_event_lock_irq(conf->wait_done,
+			bh->b_rsector < conf->start_active ||
+			bh->b_rsector >= conf->start_future,
+			conf->segment_lock);
+	if (bh->b_rsector < conf->start_active) 
+		conf->cnt_done++;
+	else {
+		conf->cnt_future++;
+		if (conf->phase)
+			set_bit(MPBH_SyncPhase, &r1_bh->state);
+	}
+	spin_unlock_irq(&conf->segment_lock);
+	
+	/*
+	 * i think the read and write branch should be separated completely,
+	 * since we want to do read balancing on the read side for example.
+	 * Alternative implementations? :) --mingo
+	 */
+
+	r1_bh->master_bh = bh;
+	r1_bh->mddev = mddev;
+	r1_bh->cmd = rw;
+
+	/*
+	 * read balancing logic:
+	 */
+	multipath = conf->multipaths + multipath_read_balance(conf);
+
+	bh_req = &r1_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 = r1_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;
+
+//			MD_BUG();
+			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));
+//				MD_BUG();
+			}
+			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--;
+//				MD_BUG();
+			}
+		}
+	}
+	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;
+		sdisk->write_only = 1;
+		break;
+	/*
+	 * Deactivate a spare disk:
+	 */
+	case DISKOP_SPARE_INACTIVE:
+		sdisk = conf->multipaths + spare_disk;
+		sdisk->operational = 0;
+		sdisk->write_only = 0;
+		break;
+	/*
+	 * Activate (mark read-write) the (now sync) spare disk,
+	 * which means we switch it's 'raid position' (->raid_disk)
+	 * with the failed disk. (only the first 'conf->nr_disks'
+	 * slots are used for 'real' disks and we must preserve this
+	 * property)
+	 */
+	case DISKOP_SPARE_ACTIVE:
+		sdisk = conf->multipaths + spare_disk;
+		fdisk = conf->multipaths + failed_disk;
+
+		spare_desc = &sb->disks[sdisk->number];
+		failed_desc = &sb->disks[fdisk->number];
+
+		if (spare_desc != *d) {
+			MD_BUG();
+			err = 1;
+			goto abort;
+		}
+
+		if (spare_desc->raid_disk != sdisk->raid_disk) {
+			MD_BUG();
+			err = 1;
+			goto abort;
+		}
+			
+		if (sdisk->raid_disk != spare_disk) {
+			MD_BUG();
+			err = 1;
+			goto abort;
+		}
+
+		if (failed_desc->raid_disk != fdisk->raid_disk) {
+			MD_BUG();
+			err = 1;
+			goto abort;
+		}
+
+		if (fdisk->raid_disk != failed_disk) {
+			MD_BUG();
+			err = 1;
+			goto abort;
+		}
+
+		/*
+		 * do the switch finally
+		 */
+		spare_rdev = find_rdev_nr(mddev, spare_desc->number);
+		failed_rdev = find_rdev_nr(mddev, failed_desc->number);
+		xchg_values(spare_rdev->desc_nr, failed_rdev->desc_nr);
+//		if (failed_rdev->alias_device)
+//			MD_BUG();
+//		if (!spare_rdev->alias_device)
+//			MD_BUG();
+		spare_rdev->alias_device = 0;
+		failed_rdev->alias_device = 1;
+
+		xchg_values(*spare_desc, *failed_desc);
+		xchg_values(*fdisk, *sdisk);
+
+		/*
+		 * (careful, 'failed' and 'spare' are switched from now on)
+		 *
+		 * we want to preserve linear numbering and we want to
+		 * give the proper raid_disk number to the now activated
+		 * disk. (this means we switch back these values)
+		 */
+	
+		xchg_values(spare_desc->raid_disk, failed_desc->raid_disk);
+		xchg_values(sdisk->raid_disk, fdisk->raid_disk);
+		xchg_values(spare_desc->number, failed_desc->number);
+		xchg_values(sdisk->number, fdisk->number);
+
+		*d = failed_desc;
+
+		if (sdisk->dev == MKDEV(0,0))
+			sdisk->used_slot = 0;
+		/*
+		 * this really activates the spare.
+		 */
+		fdisk->spare = 0;
+		fdisk->write_only = 0;
+
+		/*
+		 * if we activate a spare, we definitely replace a
+		 * non-operational disk slot in the 'low' area of
+		 * the disk array.
+		 */
+
+		conf->working_disks++;
+
+		break;
+
+	case DISKOP_HOT_REMOVE_DISK:
+		rdisk = conf->multipaths + removed_disk;
+
+		if (rdisk->spare && (removed_disk < conf->raid_disks)) {
+			MD_BUG();	
+			err = 1;
+			goto abort;
+		}
+		rdisk->dev = MKDEV(0,0);
+		rdisk->used_slot = 0;
+		conf->nr_disks--;
+		break;
+
+	case DISKOP_HOT_ADD_DISK:
+		adisk = conf->multipaths + added_disk;
+		added_desc = *d;
+
+		if (added_disk != added_desc->number) {
+			MD_BUG();	
+			err = 1;
+			goto abort;
+		}
+
+		adisk->number = added_desc->number;
+		adisk->raid_disk = added_desc->raid_disk;
+		adisk->dev = MKDEV(added_desc->major,added_desc->minor);
+
+		adisk->operational = 0;
+		adisk->write_only = 0;
+		adisk->spare = 1;
+		adisk->used_slot = 1;
+		adisk->head_position = 0;
+		conf->nr_disks++;
+
+		break;
+
+	default:
+		MD_BUG();	
+		err = 1;
+		goto abort;
+	}
+abort:
+	md_spin_unlock_irq(&conf->device_lock);
+
+	print_multipath_conf(conf);
+	return err;
+}
+
+
+#define IO_ERROR KERN_ALERT \
+"multipath: %s: unrecoverable IO read error for block %lu\n"
+
+#define REDIRECT_SECTOR KERN_ERR \
+"multipath: %s: redirecting sector %lu to another IO path\n"
+
+/*
+ * 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 (void *data)
+{
+	struct multipath_bh *r1_bh;
+	struct buffer_head *bh;
+	unsigned long flags;
+	mddev_t *mddev;
+	kdev_t dev;
+
+
+	for (;;) {
+		md_spin_lock_irqsave(&retry_list_lock, flags);
+		r1_bh = multipath_retry_list;
+		if (!r1_bh)
+			break;
+		multipath_retry_list = r1_bh->next_r1;
+		md_spin_unlock_irqrestore(&retry_list_lock, flags);
+
+		mddev = r1_bh->mddev;
+		if (mddev->sb_dirty) {
+			printk(KERN_INFO "dirty sb detected, updating.\n");
+			mddev->sb_dirty = 0;
+			md_update_sb(mddev);
+		}
+		bh = &r1_bh->bh_req;
+		dev = bh->b_dev;
+		
+		multipath_map (mddev, &bh->b_dev);
+		if (bh->b_dev == dev) {
+			printk (IO_ERROR, partition_name(bh->b_dev), bh->b_blocknr);
+			multipath_end_bh_io(r1_bh, 0);
+		} else {
+			printk (REDIRECT_SECTOR,
+				partition_name(bh->b_dev), bh->b_blocknr);
+			bh->b_rdev = bh->b_dev;
+			bh->b_rsector = bh->b_blocknr;
+			generic_make_request (r1_bh->cmd, bh);
+		}
+	}
+	md_spin_unlock_irqrestore(&retry_list_lock, flags);
+}
+#undef IO_ERROR
+#undef REDIRECT_SECTOR
+
+/*
+ * This will catch the scenario in which one of the multipaths was
+ * mounted as a normal device rather than as a part of a raid set.
+ *
+ * check_consistency is very personality-dependent, eg. RAID5 cannot
+ * do this check, it uses another method.
+ */
+static int __check_consistency (mddev_t *mddev, int row)
+{
+	multipath_conf_t *conf = mddev_to_conf(mddev);
+	int disks = MD_SB_DISKS;
+	kdev_t dev;
+	struct buffer_head *bh = NULL;
+	int i, rc = 0;
+	char *buffer = NULL;
+
+	for (i = 0; i < disks; i++) {
+		if (!conf->multipaths[i].operational)
+			continue;
+		printk("(checking disk %d)\n",i);
+		dev = conf->multipaths[i].dev;
+		set_blocksize(dev, 4096);
+		if ((bh = bread(dev, row / 4, 4096)) == NULL)
+			break;
+		if (!buffer) {
+			buffer = (char *) __get_free_page(GFP_KERNEL);
+			if (!buffer)
+				break;
+			memcpy(buffer, bh->b_data, 4096);
+		} else if (memcmp(buffer, bh->b_data, 4096)) {
+			rc = 1;
+			break;
+		}
+		bforget(bh);
+		fsync_dev(dev);
+		invalidate_buffers(dev);
+		bh = NULL;
+	}
+	if (buffer)
+		free_page((unsigned long) buffer);
+	if (bh) {
+		dev = bh->b_dev;
+		bforget(bh);
+		fsync_dev(dev);
+		invalidate_buffers(dev);
+	}
+	return rc;
+}
+
+static int check_consistency (mddev_t *mddev)
+{
+	if (__check_consistency(mddev, 0))
+/*
+ * we do not do this currently, as it's perfectly possible to
+ * have an inconsistent array when it's freshly created. Only
+ * newly written data has to be consistent.
+ */
+		return 0;
+
+	return 0;
+}
+
+#define INVALID_LEVEL KERN_WARNING \
+"multipath: md%d: raid level not set to multipath IO (%d)\n"
+
+#define NO_SB KERN_ERR \
+"multipath: disabled IO path %s (couldn't access raid superblock)\n"
+
+#define ERRORS KERN_ERR \
+"multipath: disabled IO path %s (errors detected)\n"
+
+#define NOT_IN_SYNC KERN_ERR \
+"multipath: making IO path %s a spare path (not in sync)\n"
+
+#define INCONSISTENT KERN_ERR \
+"multipath: disabled IO path %s (inconsistent descriptor)\n"
+
+#define ALREADY_RUNNING KERN_ERR \
+"multipath: disabled IO path %s (multipath %d already operational)\n"
+
+#define OPERATIONAL KERN_INFO \
+"multipath: device %s operational as IO path %d\n"
+
+#define MEM_ERROR KERN_ERR \
+"multipath: couldn't allocate memory for md%d\n"
+
+#define SPARE KERN_INFO \
+"multipath: spare IO path %s\n"
+
+#define NONE_OPERATIONAL KERN_ERR \
+"multipath: no operational IO paths for md%d\n"
+
+#define SB_DIFFERENCES KERN_ERR \
+"multipath: detected IO path differences!\n"
+
+#define ARRAY_IS_ACTIVE KERN_INFO \
+"multipath: array md%d active with %d out of %d IO paths (%d spare IO paths)\n"
+
+#define THREAD_ERROR KERN_ERR \
+"multipath: couldn't allocate thread for md%d\n"
+
+static int multipath_run (mddev_t *mddev)
+{
+	multipath_conf_t *conf;
+	int i, j, disk_idx;
+	struct multipath_info *disk, *disk2;
+	mdp_super_t *sb = mddev->sb;
+	mdp_disk_t *desc, *desc2;
+	mdk_rdev_t *rdev, *def_rdev = NULL;
+	struct md_list_head *tmp;
+	int start_recovery = 0, num_rdevs = 0;
+
+	MOD_INC_USE_COUNT;
+
+	if (sb->level != -4) {
+		printk(INVALID_LEVEL, mdidx(mddev), sb->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()]
+	 */
+
+	conf = kmalloc(sizeof(multipath_conf_t), GFP_KERNEL);
+	mddev->private = conf;
+	if (!conf) {
+		printk(MEM_ERROR, mdidx(mddev));
+		goto out;
+	}
+	memset(conf, 0, sizeof(*conf));
+
+	ITERATE_RDEV(mddev,rdev,tmp) {
+		if (rdev->faulty) {
+			/* this is a "should never happen" case and if it */
+			/* ever does happen, a continue; won't help */
+			printk(ERRORS, partition_name(rdev->dev));
+			continue;
+		} else {
+			/* this is a "should never happen" case and if it */
+			/* ever does happen, a continue; won't help */
+			if (!rdev->sb) {
+				MD_BUG();
+				continue;
+			}
+		}
+		if (rdev->desc_nr == -1) {
+			MD_BUG();
+			continue;
+		}
+
+		desc = &sb->disks[rdev->desc_nr];
+		disk_idx = desc->raid_disk;
+		disk = conf->multipaths + disk_idx;
+
+		if (!disk_sync(desc))
+			printk(NOT_IN_SYNC, partition_name(rdev->dev));
+
+		/*
+		 * Mark all disks as spare to start with, then pick our
+		 * active disk.  If we have a disk that is marked active
+		 * in the sb, then use it, else use the first rdev.
+		 */
+		disk->number = desc->number;
+		disk->raid_disk = desc->raid_disk;
+		disk->dev = rdev->dev;
+		disk->sect_limit = MAX_WORK_PER_DISK;
+		disk->operational = 0;
+		disk->write_only = 0;
+		disk->spare = 1;
+		disk->used_slot = 1;
+		disk->head_position = 0;
+		mark_disk_sync(desc);
+
+		if (disk_active(desc)) {
+			if(!conf->working_disks) {
+				printk(OPERATIONAL, partition_name(rdev->dev),
+ 					desc->raid_disk);
+				disk->operational = 1;
+				disk->spare = 0;
+				conf->working_disks++;
+				def_rdev = rdev;
+			} else {
+				mark_disk_spare(desc);
+			}
+		} else
+			mark_disk_spare(desc);
+
+		if(!num_rdevs++) def_rdev = rdev;
+	}
+	if(!conf->working_disks && num_rdevs) {
+		desc = &sb->disks[def_rdev->desc_nr];
+		disk = conf->multipaths + desc->raid_disk;
+		printk(OPERATIONAL, partition_name(def_rdev->dev),
+			disk->raid_disk);
+		disk->operational = 1;
+		disk->spare = 0;
+		conf->working_disks++;
+		mark_disk_active(desc);
+	}
+	/*
+	 * Make sure our active path is in desc spot 0
+	 */
+	if(def_rdev->desc_nr != 0) {
+		rdev = find_rdev_nr(mddev, 0);
+		desc = &sb->disks[def_rdev->desc_nr];
+		desc2 = sb->disks;
+		disk = conf->multipaths + desc->raid_disk;
+		disk2 = conf->multipaths + desc2->raid_disk;
+		xchg_values(*desc2,*desc);
+		xchg_values(*disk2,*disk);
+		xchg_values(desc2->number, desc->number);
+		xchg_values(disk2->number, disk->number);
+		xchg_values(desc2->raid_disk, desc->raid_disk);
+		xchg_values(disk2->raid_disk, disk->raid_disk);
+		if(rdev) {
+			xchg_values(def_rdev->desc_nr,rdev->desc_nr);
+		} else {
+			def_rdev->desc_nr = 0;
+		}
+	}
+	conf->raid_disks = sb->raid_disks = sb->active_disks = 1;
+	conf->nr_disks = sb->nr_disks = sb->working_disks = num_rdevs;
+	sb->failed_disks = 0;
+	sb->spare_disks = num_rdevs - 1;
+	mddev->sb_dirty = 1;
+	conf->mddev = mddev;
+	conf->device_lock = MD_SPIN_LOCK_UNLOCKED;
+
+	conf->segment_lock = MD_SPIN_LOCK_UNLOCKED;
+	init_waitqueue_head(&conf->wait_buffer);
+	init_waitqueue_head(&conf->wait_done);
+	init_waitqueue_head(&conf->wait_ready);
+
+	if (!conf->working_disks) {
+		printk(NONE_OPERATIONAL, mdidx(mddev));
+		goto out_free_conf;
+	}
+
+
+	/* pre-allocate some buffer_head structures.
+	 * As a minimum, 1 mpbh and raid_disks buffer_heads
+	 * would probably get us by in tight memory situations,
+	 * but a few more is probably a good idea.
+	 * For now, try 16 mpbh and 16*raid_disks bufferheads
+	 * This will allow at least 16 concurrent reads or writes
+	 * even if kmalloc starts failing
+	 */
+	if (multipath_grow_mpbh(conf, 16) < 16 ||
+	    multipath_grow_bh(conf, 16*conf->raid_disks)< 16*conf->raid_disks) {
+		printk(MEM_ERROR, mdidx(mddev));
+		goto out_free_conf;
+	}
+
+	if (!start_recovery && (sb->state & (1 << MD_SB_CLEAN))) {
+		/*
+		 * we do sanity checks even if the device says
+		 * it's clean ...
+		 */
+		if (check_consistency(mddev)) {
+			printk(SB_DIFFERENCES);
+			sb->state &= ~(1 << MD_SB_CLEAN);
+		}
+	}
+
+	{
+		const char * name = "multipathd";
+
+		conf->thread = md_register_thread(multipathd, conf, name);
+		if (!conf->thread) {
+			printk(THREAD_ERROR, mdidx(mddev));
+			goto out_free_conf;
+		}
+	}
+
+	/*
+	 * Regenerate the "device is in sync with the raid set" bit for
+	 * each device.
+	 */
+	for (i = 0; i < MD_SB_DISKS; i++) {
+		mark_disk_nonsync(sb->disks+i);
+		for (j = 0; j < sb->raid_disks; j++) {
+			if (sb->disks[i].number == conf->multipaths[j].number)
+				mark_disk_sync(sb->disks+i);
+		}
+	}
+
+	printk(ARRAY_IS_ACTIVE, mdidx(mddev), sb->active_disks,
+			sb->raid_disks, sb->spare_disks);
+	/*
+	 * Ok, everything is just fine now
+	 */
+	return 0;
+
+out_free_conf:
+	multipath_shrink_mpbh(conf);
+	multipath_shrink_bh(conf, conf->freebh_cnt);
+	kfree(conf);
+	mddev->private = NULL;
+out:
+	MOD_DEC_USE_COUNT;
+	return -EIO;
+}
+
+#undef INVALID_LEVEL
+#undef NO_SB
+#undef ERRORS
+#undef NOT_IN_SYNC
+#undef INCONSISTENT
+#undef ALREADY_RUNNING
+#undef OPERATIONAL
+#undef SPARE
+#undef NONE_OPERATIONAL
+#undef SB_DIFFERENCES
+#undef ARRAY_IS_ACTIVE
+
+static int multipath_stop (mddev_t *mddev)
+{
+	multipath_conf_t *conf = mddev_to_conf(mddev);
+
+	md_unregister_thread(conf->thread);
+	multipath_shrink_mpbh(conf);
+	multipath_shrink_bh(conf, conf->freebh_cnt);
+	kfree(conf);
+	mddev->private = NULL;
+	MOD_DEC_USE_COUNT;
+	return 0;
+}
+
+static mdk_personality_t multipath_personality=
+{
+	name:		"multipath",
+	make_request:	multipath_make_request,
+	run:		multipath_run,
+	stop:		multipath_stop,
+	status:		multipath_status,
+	error_handler:	multipath_error,
+	diskop:		multipath_diskop,
+};
+
+static int md__init multipath_init (void)
+{
+	return register_md_personality (MULTIPATH, &multipath_personality);
+}
+
+static void multipath_exit (void)
+{
+	unregister_md_personality (MULTIPATH);
+}
+
+module_init(multipath_init);
+module_exit(multipath_exit);
+

FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen (who was at: slshen@lbl.gov)