X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fmd%2Fdm-mpath.c;h=dce971dbdfa3a4f034d5934c5950de8501e57b29;hb=c0da3748b9a894b9f9b561ecc2d090a913988a0f;hp=84cdb700a247040d4e72a7f79c03df12a7598ced;hpb=f6a80ea8ed44de0b19c42d41928be37a186a3f41;p=safe%2Fjmp%2Flinux-2.6

diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 84cdb70..dce971d 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -5,11 +5,10 @@
  * This file is released under the GPL.
  */
 
-#include "dm.h"
+#include <linux/device-mapper.h>
+
 #include "dm-path-selector.h"
-#include "dm-hw-handler.h"
-#include "dm-bio-list.h"
-#include "dm-bio-record.h"
+#include "dm-uevent.h"
 
 #include <linux/ctype.h>
 #include <linux/init.h>
@@ -19,8 +18,10 @@
 #include <linux/slab.h>
 #include <linux/time.h>
 #include <linux/workqueue.h>
+#include <scsi/scsi_dh.h>
 #include <asm/atomic.h>
 
+#define DM_MSG_PREFIX "multipath"
 #define MESG_STR(x) x, sizeof(x)
 
 /* Path properties */
@@ -28,9 +29,12 @@ struct pgpath {
 	struct list_head list;
 
 	struct priority_group *pg;	/* Owning PG */
+	unsigned is_active;		/* Path status */
 	unsigned fail_count;		/* Cumulative failure count */
 
-	struct path path;
+	struct dm_path path;
+	struct work_struct deactivate_path;
+	struct work_struct activate_path;
 };
 
 #define path_to_pgpath(__pgp) container_of((__pgp), struct pgpath, path)
@@ -59,10 +63,12 @@ struct multipath {
 
 	spinlock_t lock;
 
-	struct hw_handler hw_handler;
+	const char *hw_handler_name;
+	char *hw_handler_params;
 	unsigned nr_priority_groups;
 	struct list_head priority_groups;
 	unsigned pg_init_required;	/* pg_init needs calling? */
+	unsigned pg_init_in_progress;	/* Only one pg_init allowed at once */
 
 	unsigned nr_valid_paths;	/* Total number of usable paths */
 	struct pgpath *current_pgpath;
@@ -72,16 +78,18 @@ struct multipath {
 
 	unsigned queue_io;		/* Must we queue all I/O? */
 	unsigned queue_if_no_path;	/* Queue I/O if last path fails? */
-	unsigned suspended;		/* Has dm core suspended our I/O? */
+	unsigned saved_queue_if_no_path;/* Saved state during suspension */
+	unsigned pg_init_retries;	/* Number of times to retry pg_init */
+	unsigned pg_init_count;		/* Number of times pg_init called */
 
 	struct work_struct process_queued_ios;
-	struct bio_list queued_ios;
+	struct list_head queued_ios;
 	unsigned queue_size;
 
 	struct work_struct trigger_event;
 
 	/*
-	 * We must use a mempool of mpath_io structs so that we
+	 * We must use a mempool of dm_mpath_io structs so that we
 	 * can resubmit bios on error.
 	 */
 	mempool_t *mpio_pool;
@@ -90,20 +98,22 @@ struct multipath {
 /*
  * Context information attached to each bio we process.
  */
-struct mpath_io {
+struct dm_mpath_io {
 	struct pgpath *pgpath;
-	struct dm_bio_details details;
+	size_t nr_bytes;
 };
 
 typedef int (*action_fn) (struct pgpath *pgpath);
 
 #define MIN_IOS 256	/* Mempool size */
 
-static kmem_cache_t *_mpio_cache;
+static struct kmem_cache *_mpio_cache;
 
-struct workqueue_struct *kmultipathd;
-static void process_queued_ios(void *data);
-static void trigger_event(void *data);
+static struct workqueue_struct *kmultipathd, *kmpath_handlerd;
+static void process_queued_ios(struct work_struct *work);
+static void trigger_event(struct work_struct *work);
+static void activate_path(struct work_struct *work);
+static void deactivate_path(struct work_struct *work);
 
 
 /*-----------------------------------------------
@@ -112,31 +122,38 @@ static void trigger_event(void *data);
 
 static struct pgpath *alloc_pgpath(void)
 {
-	struct pgpath *pgpath = kmalloc(sizeof(*pgpath), GFP_KERNEL);
+	struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL);
 
 	if (pgpath) {
-		memset(pgpath, 0, sizeof(*pgpath));
-		pgpath->path.is_active = 1;
+		pgpath->is_active = 1;
+		INIT_WORK(&pgpath->deactivate_path, deactivate_path);
+		INIT_WORK(&pgpath->activate_path, activate_path);
 	}
 
 	return pgpath;
 }
 
-static inline void free_pgpath(struct pgpath *pgpath)
+static void free_pgpath(struct pgpath *pgpath)
 {
 	kfree(pgpath);
 }
 
+static void deactivate_path(struct work_struct *work)
+{
+	struct pgpath *pgpath =
+		container_of(work, struct pgpath, deactivate_path);
+
+	blk_abort_queue(pgpath->path.dev->bdev->bd_disk->queue);
+}
+
 static struct priority_group *alloc_priority_group(void)
 {
 	struct priority_group *pg;
 
-	pg = kmalloc(sizeof(*pg), GFP_KERNEL);
-	if (!pg)
-		return NULL;
+	pg = kzalloc(sizeof(*pg), GFP_KERNEL);
 
-	memset(pg, 0, sizeof(*pg));
-	INIT_LIST_HEAD(&pg->pgpaths);
+	if (pg)
+		INIT_LIST_HEAD(&pg->pgpaths);
 
 	return pg;
 }
@@ -144,9 +161,12 @@ static struct priority_group *alloc_priority_group(void)
 static void free_pgpaths(struct list_head *pgpaths, struct dm_target *ti)
 {
 	struct pgpath *pgpath, *tmp;
+	struct multipath *m = ti->private;
 
 	list_for_each_entry_safe(pgpath, tmp, pgpaths, list) {
 		list_del(&pgpath->list);
+		if (m->hw_handler_name)
+			scsi_dh_detach(bdev_get_queue(pgpath->path.dev->bdev));
 		dm_put_device(ti, pgpath->path.dev);
 		free_pgpath(pgpath);
 	}
@@ -166,24 +186,25 @@ static void free_priority_group(struct priority_group *pg,
 	kfree(pg);
 }
 
-static struct multipath *alloc_multipath(void)
+static struct multipath *alloc_multipath(struct dm_target *ti)
 {
 	struct multipath *m;
 
-	m = kmalloc(sizeof(*m), GFP_KERNEL);
+	m = kzalloc(sizeof(*m), GFP_KERNEL);
 	if (m) {
-		memset(m, 0, sizeof(*m));
 		INIT_LIST_HEAD(&m->priority_groups);
+		INIT_LIST_HEAD(&m->queued_ios);
 		spin_lock_init(&m->lock);
 		m->queue_io = 1;
-		INIT_WORK(&m->process_queued_ios, process_queued_ios, m);
-		INIT_WORK(&m->trigger_event, trigger_event, m);
-		m->mpio_pool = mempool_create(MIN_IOS, mempool_alloc_slab,
-					      mempool_free_slab, _mpio_cache);
+		INIT_WORK(&m->process_queued_ios, process_queued_ios);
+		INIT_WORK(&m->trigger_event, trigger_event);
+		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
 		if (!m->mpio_pool) {
 			kfree(m);
 			return NULL;
 		}
+		m->ti = ti;
+		ti->private = m;
 	}
 
 	return m;
@@ -192,18 +213,14 @@ static struct multipath *alloc_multipath(void)
 static void free_multipath(struct multipath *m)
 {
 	struct priority_group *pg, *tmp;
-	struct hw_handler *hwh = &m->hw_handler;
 
 	list_for_each_entry_safe(pg, tmp, &m->priority_groups, list) {
 		list_del(&pg->list);
 		free_priority_group(pg, m->ti);
 	}
 
-	if (hwh->type) {
-		hwh->type->destroy(hwh);
-		dm_put_hw_handler(hwh->type);
-	}
-
+	kfree(m->hw_handler_name);
+	kfree(m->hw_handler_params);
 	mempool_destroy(m->mpio_pool);
 	kfree(m);
 }
@@ -215,25 +232,26 @@ static void free_multipath(struct multipath *m)
 
 static void __switch_pg(struct multipath *m, struct pgpath *pgpath)
 {
-	struct hw_handler *hwh = &m->hw_handler;
-
 	m->current_pg = pgpath->pg;
 
 	/* Must we initialise the PG first, and queue I/O till it's ready? */
-	if (hwh->type && hwh->type->pg_init) {
+	if (m->hw_handler_name) {
 		m->pg_init_required = 1;
 		m->queue_io = 1;
 	} else {
 		m->pg_init_required = 0;
 		m->queue_io = 0;
 	}
+
+	m->pg_init_count = 0;
 }
 
-static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg)
+static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg,
+			       size_t nr_bytes)
 {
-	struct path *path;
+	struct dm_path *path;
 
-	path = pg->ps.type->select_path(&pg->ps, &m->repeat_count);
+	path = pg->ps.type->select_path(&pg->ps, &m->repeat_count, nr_bytes);
 	if (!path)
 		return -ENXIO;
 
@@ -245,7 +263,7 @@ static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg)
 	return 0;
 }
 
-static void __choose_pgpath(struct multipath *m)
+static void __choose_pgpath(struct multipath *m, size_t nr_bytes)
 {
 	struct priority_group *pg;
 	unsigned bypassed = 1;
@@ -257,12 +275,12 @@ static void __choose_pgpath(struct multipath *m)
 	if (m->next_pg) {
 		pg = m->next_pg;
 		m->next_pg = NULL;
-		if (!__choose_path_in_pg(m, pg))
+		if (!__choose_path_in_pg(m, pg, nr_bytes))
 			return;
 	}
 
 	/* Don't change PG until it has no remaining paths */
-	if (m->current_pg && !__choose_path_in_pg(m, m->current_pg))
+	if (m->current_pg && !__choose_path_in_pg(m, m->current_pg, nr_bytes))
 		return;
 
 	/*
@@ -274,7 +292,7 @@ static void __choose_pgpath(struct multipath *m)
 		list_for_each_entry(pg, &m->priority_groups, list) {
 			if (pg->bypassed == bypassed)
 				continue;
-			if (!__choose_path_in_pg(m, pg))
+			if (!__choose_path_in_pg(m, pg, nr_bytes))
 				return;
 		}
 	} while (bypassed--);
@@ -284,19 +302,38 @@ failed:
 	m->current_pg = NULL;
 }
 
-static int map_io(struct multipath *m, struct bio *bio, struct mpath_io *mpio,
-		  unsigned was_queued)
+/*
+ * Check whether bios must be queued in the device-mapper core rather
+ * than here in the target.
+ *
+ * m->lock must be held on entry.
+ *
+ * If m->queue_if_no_path and m->saved_queue_if_no_path hold the
+ * same value then we are not between multipath_presuspend()
+ * and multipath_resume() calls and we have no need to check
+ * for the DMF_NOFLUSH_SUSPENDING flag.
+ */
+static int __must_push_back(struct multipath *m)
+{
+	return (m->queue_if_no_path != m->saved_queue_if_no_path &&
+		dm_noflush_suspending(m->ti));
+}
+
+static int map_io(struct multipath *m, struct request *clone,
+		  struct dm_mpath_io *mpio, unsigned was_queued)
 {
-	int r = 1;
+	int r = DM_MAPIO_REMAPPED;
+	size_t nr_bytes = blk_rq_bytes(clone);
 	unsigned long flags;
 	struct pgpath *pgpath;
+	struct block_device *bdev;
 
 	spin_lock_irqsave(&m->lock, flags);
 
 	/* Do we need to select a new pgpath? */
 	if (!m->current_pgpath ||
 	    (!m->queue_io && (m->repeat_count && --m->repeat_count == 0)))
-		__choose_pgpath(m);
+		__choose_pgpath(m, nr_bytes);
 
 	pgpath = m->current_pgpath;
 
@@ -304,20 +341,30 @@ static int map_io(struct multipath *m, struct bio *bio, struct mpath_io *mpio,
 		m->queue_size--;
 
 	if ((pgpath && m->queue_io) ||
-	    (!pgpath && m->queue_if_no_path && !m->suspended)) {
+	    (!pgpath && m->queue_if_no_path)) {
 		/* Queue for the daemon to resubmit */
-		bio_list_add(&m->queued_ios, bio);
+		list_add_tail(&clone->queuelist, &m->queued_ios);
 		m->queue_size++;
-		if (m->pg_init_required || !m->queue_io)
+		if ((m->pg_init_required && !m->pg_init_in_progress) ||
+		    !m->queue_io)
 			queue_work(kmultipathd, &m->process_queued_ios);
 		pgpath = NULL;
-		r = 0;
-	} else if (!pgpath)
-		r = -EIO;		/* Failed */
+		r = DM_MAPIO_SUBMITTED;
+	} else if (pgpath) {
+		bdev = pgpath->path.dev->bdev;
+		clone->q = bdev_get_queue(bdev);
+		clone->rq_disk = bdev->bd_disk;
+	} else if (__must_push_back(m))
+		r = DM_MAPIO_REQUEUE;
 	else
-		bio->bi_bdev = pgpath->path.dev->bdev;
+		r = -EIO;	/* Failed */
 
 	mpio->pgpath = pgpath;
+	mpio->nr_bytes = nr_bytes;
+
+	if (r == DM_MAPIO_REMAPPED && pgpath->pg->ps.type->start_io)
+		pgpath->pg->ps.type->start_io(&pgpath->pg->ps, &pgpath->path,
+					      nr_bytes);
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
@@ -327,14 +374,19 @@ static int map_io(struct multipath *m, struct bio *bio, struct mpath_io *mpio,
 /*
  * If we run out of usable paths, should we queue I/O or error it?
  */
-static int queue_if_no_path(struct multipath *m, unsigned queue_if_no_path)
+static int queue_if_no_path(struct multipath *m, unsigned queue_if_no_path,
+			    unsigned save_old_value)
 {
 	unsigned long flags;
 
 	spin_lock_irqsave(&m->lock, flags);
 
+	if (save_old_value)
+		m->saved_queue_if_no_path = m->queue_if_no_path;
+	else
+		m->saved_queue_if_no_path = queue_if_no_path;
 	m->queue_if_no_path = queue_if_no_path;
-	if (!m->queue_if_no_path)
+	if (!m->queue_if_no_path && m->queue_size)
 		queue_work(kmultipathd, &m->process_queued_ios);
 
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -350,59 +402,66 @@ static void dispatch_queued_ios(struct multipath *m)
 {
 	int r;
 	unsigned long flags;
-	struct bio *bio = NULL, *next;
-	struct mpath_io *mpio;
+	struct dm_mpath_io *mpio;
 	union map_info *info;
+	struct request *clone, *n;
+	LIST_HEAD(cl);
 
 	spin_lock_irqsave(&m->lock, flags);
-	bio = bio_list_get(&m->queued_ios);
+	list_splice_init(&m->queued_ios, &cl);
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	while (bio) {
-		next = bio->bi_next;
-		bio->bi_next = NULL;
+	list_for_each_entry_safe(clone, n, &cl, queuelist) {
+		list_del_init(&clone->queuelist);
 
-		info = dm_get_mapinfo(bio);
+		info = dm_get_rq_mapinfo(clone);
 		mpio = info->ptr;
 
-		r = map_io(m, bio, mpio, 1);
-		if (r < 0)
-			bio_endio(bio, bio->bi_size, r);
-		else if (r == 1)
-			generic_make_request(bio);
-
-		bio = next;
+		r = map_io(m, clone, mpio, 1);
+		if (r < 0) {
+			mempool_free(mpio, m->mpio_pool);
+			dm_kill_unmapped_request(clone, r);
+		} else if (r == DM_MAPIO_REMAPPED)
+			dm_dispatch_request(clone);
+		else if (r == DM_MAPIO_REQUEUE) {
+			mempool_free(mpio, m->mpio_pool);
+			dm_requeue_unmapped_request(clone);
+		}
 	}
 }
 
-static void process_queued_ios(void *data)
+static void process_queued_ios(struct work_struct *work)
 {
-	struct multipath *m = (struct multipath *) data;
-	struct hw_handler *hwh = &m->hw_handler;
-	struct pgpath *pgpath;
-	unsigned init_required, must_queue = 0;
+	struct multipath *m =
+		container_of(work, struct multipath, process_queued_ios);
+	struct pgpath *pgpath = NULL, *tmp;
+	unsigned must_queue = 1;
 	unsigned long flags;
 
 	spin_lock_irqsave(&m->lock, flags);
 
+	if (!m->queue_size)
+		goto out;
+
 	if (!m->current_pgpath)
-		__choose_pgpath(m);
+		__choose_pgpath(m, 0);
 
 	pgpath = m->current_pgpath;
 
-	if ((pgpath && m->queue_io) ||
-	    (!pgpath && m->queue_if_no_path && !m->suspended))
-		must_queue = 1;
+	if ((pgpath && !m->queue_io) ||
+	    (!pgpath && !m->queue_if_no_path))
+		must_queue = 0;
 
-	init_required = m->pg_init_required;
-	if (init_required)
+	if (m->pg_init_required && !m->pg_init_in_progress && pgpath) {
+		m->pg_init_count++;
 		m->pg_init_required = 0;
-
+		list_for_each_entry(tmp, &pgpath->pg->pgpaths, list) {
+			if (queue_work(kmpath_handlerd, &tmp->activate_path))
+				m->pg_init_in_progress++;
+		}
+	}
+out:
 	spin_unlock_irqrestore(&m->lock, flags);
-
-	if (init_required)
-		hwh->type->pg_init(hwh, pgpath->pg->bypassed, &pgpath->path);
-
 	if (!must_queue)
 		dispatch_queued_ios(m);
 }
@@ -411,9 +470,10 @@ static void process_queued_ios(void *data)
  * An event is triggered whenever a path is taken out of use.
  * Includes path failure and PG bypass.
  */
-static void trigger_event(void *data)
+static void trigger_event(struct work_struct *work)
 {
-	struct multipath *m = (struct multipath *) data;
+	struct multipath *m =
+		container_of(work, struct multipath, trigger_event);
 
 	dm_table_event(m->ti->table);
 }
@@ -434,8 +494,6 @@ struct param {
 	char *error;
 };
 
-#define ESTR(s) ("dm-multipath: " s)
-
 static int read_param(struct param *param, char *str, unsigned *v, char **error)
 {
 	if (!str ||
@@ -483,23 +541,31 @@ static int parse_path_selector(struct arg_set *as, struct priority_group *pg,
 	unsigned ps_argc;
 
 	static struct param _params[] = {
-		{0, 1024, ESTR("invalid number of path selector args")},
+		{0, 1024, "invalid number of path selector args"},
 	};
 
 	pst = dm_get_path_selector(shift(as));
 	if (!pst) {
-		ti->error = ESTR("unknown path selector type");
+		ti->error = "unknown path selector type";
 		return -EINVAL;
 	}
 
 	r = read_param(_params, shift(as), &ps_argc, &ti->error);
-	if (r)
+	if (r) {
+		dm_put_path_selector(pst);
+		return -EINVAL;
+	}
+
+	if (ps_argc > as->argc) {
+		dm_put_path_selector(pst);
+		ti->error = "not enough arguments for path selector";
 		return -EINVAL;
+	}
 
 	r = pst->create(&pg->ps, ps_argc, as->argv);
 	if (r) {
 		dm_put_path_selector(pst);
-		ti->error = ESTR("path selector constructor failed");
+		ti->error = "path selector constructor failed";
 		return r;
 	}
 
@@ -514,24 +580,56 @@ static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
 {
 	int r;
 	struct pgpath *p;
+	struct multipath *m = ti->private;
 
 	/* we need at least a path arg */
 	if (as->argc < 1) {
-		ti->error = ESTR("no device given");
-		return NULL;
+		ti->error = "no device given";
+		return ERR_PTR(-EINVAL);
 	}
 
 	p = alloc_pgpath();
 	if (!p)
-		return NULL;
+		return ERR_PTR(-ENOMEM);
 
 	r = dm_get_device(ti, shift(as), ti->begin, ti->len,
 			  dm_table_get_mode(ti->table), &p->path.dev);
 	if (r) {
-		ti->error = ESTR("error getting device");
+		ti->error = "error getting device";
 		goto bad;
 	}
 
+	if (m->hw_handler_name) {
+		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
+
+		r = scsi_dh_attach(q, m->hw_handler_name);
+		if (r == -EBUSY) {
+			/*
+			 * Already attached to different hw_handler,
+			 * try to reattach with correct one.
+			 */
+			scsi_dh_detach(q);
+			r = scsi_dh_attach(q, m->hw_handler_name);
+		}
+
+		if (r < 0) {
+			ti->error = "error attaching hardware handler";
+			dm_put_device(ti, p->path.dev);
+			goto bad;
+		}
+
+		if (m->hw_handler_params) {
+			r = scsi_dh_set_params(q, m->hw_handler_params);
+			if (r < 0) {
+				ti->error = "unable to set hardware "
+							"handler parameters";
+				scsi_dh_detach(q);
+				dm_put_device(ti, p->path.dev);
+				goto bad;
+			}
+		}
+	}
+
 	r = ps->type->add_path(ps, &p->path, as->argc, as->argv, &ti->error);
 	if (r) {
 		dm_put_device(ti, p->path.dev);
@@ -542,32 +640,32 @@ static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
 
  bad:
 	free_pgpath(p);
-	return NULL;
+	return ERR_PTR(r);
 }
 
 static struct priority_group *parse_priority_group(struct arg_set *as,
-						   struct multipath *m,
-						   struct dm_target *ti)
+						   struct multipath *m)
 {
 	static struct param _params[] = {
-		{1, 1024, ESTR("invalid number of paths")},
-		{0, 1024, ESTR("invalid number of selector args")}
+		{1, 1024, "invalid number of paths"},
+		{0, 1024, "invalid number of selector args"}
 	};
 
 	int r;
 	unsigned i, nr_selector_args, nr_params;
 	struct priority_group *pg;
+	struct dm_target *ti = m->ti;
 
 	if (as->argc < 2) {
 		as->argc = 0;
-		ti->error = ESTR("not enough priority group aruments");
-		return NULL;
+		ti->error = "not enough priority group arguments";
+		return ERR_PTR(-EINVAL);
 	}
 
 	pg = alloc_priority_group();
 	if (!pg) {
-		ti->error = ESTR("couldn't allocate priority group");
-		return NULL;
+		ti->error = "couldn't allocate priority group";
+		return ERR_PTR(-ENOMEM);
 	}
 	pg->m = m;
 
@@ -591,15 +689,19 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
 		struct pgpath *pgpath;
 		struct arg_set path_args;
 
-		if (as->argc < nr_params)
+		if (as->argc < nr_params) {
+			ti->error = "not enough path parameters";
 			goto bad;
+		}
 
 		path_args.argc = nr_params;
 		path_args.argv = as->argv;
 
 		pgpath = parse_path(&path_args, &pg->ps, ti);
-		if (!pgpath)
+		if (IS_ERR(pgpath)) {
+			r = PTR_ERR(pgpath);
 			goto bad;
+		}
 
 		pgpath->pg = pg;
 		list_add_tail(&pgpath->list, &pg->pgpaths);
@@ -610,54 +712,73 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
 
  bad:
 	free_priority_group(pg, ti);
-	return NULL;
+	return ERR_PTR(r);
 }
 
-static int parse_hw_handler(struct arg_set *as, struct multipath *m,
-			    struct dm_target *ti)
+static int parse_hw_handler(struct arg_set *as, struct multipath *m)
 {
-	int r;
-	struct hw_handler_type *hwht;
 	unsigned hw_argc;
+	int ret;
+	struct dm_target *ti = m->ti;
 
 	static struct param _params[] = {
-		{0, 1024, ESTR("invalid number of hardware handler args")},
+		{0, 1024, "invalid number of hardware handler args"},
 	};
 
-	r = read_param(_params, shift(as), &hw_argc, &ti->error);
-	if (r)
+	if (read_param(_params, shift(as), &hw_argc, &ti->error))
 		return -EINVAL;
 
 	if (!hw_argc)
 		return 0;
 
-	hwht = dm_get_hw_handler(shift(as));
-	if (!hwht) {
-		ti->error = ESTR("unknown hardware handler type");
+	if (hw_argc > as->argc) {
+		ti->error = "not enough arguments for hardware handler";
 		return -EINVAL;
 	}
 
-	r = hwht->create(&m->hw_handler, hw_argc - 1, as->argv);
-	if (r) {
-		dm_put_hw_handler(hwht);
-		ti->error = ESTR("hardware handler constructor failed");
-		return r;
+	m->hw_handler_name = kstrdup(shift(as), GFP_KERNEL);
+	request_module("scsi_dh_%s", m->hw_handler_name);
+	if (scsi_dh_handler_exist(m->hw_handler_name) == 0) {
+		ti->error = "unknown hardware handler type";
+		ret = -EINVAL;
+		goto fail;
 	}
 
-	m->hw_handler.type = hwht;
+	if (hw_argc > 1) {
+		char *p;
+		int i, j, len = 4;
+
+		for (i = 0; i <= hw_argc - 2; i++)
+			len += strlen(as->argv[i]) + 1;
+		p = m->hw_handler_params = kzalloc(len, GFP_KERNEL);
+		if (!p) {
+			ti->error = "memory allocation failed";
+			ret = -ENOMEM;
+			goto fail;
+		}
+		j = sprintf(p, "%d", hw_argc - 1);
+		for (i = 0, p+=j+1; i <= hw_argc - 2; i++, p+=j+1)
+			j = sprintf(p, "%s", as->argv[i]);
+	}
 	consume(as, hw_argc - 1);
 
 	return 0;
+fail:
+	kfree(m->hw_handler_name);
+	m->hw_handler_name = NULL;
+	return ret;
 }
 
-static int parse_features(struct arg_set *as, struct multipath *m,
-			  struct dm_target *ti)
+static int parse_features(struct arg_set *as, struct multipath *m)
 {
 	int r;
 	unsigned argc;
+	struct dm_target *ti = m->ti;
+	const char *param_name;
 
 	static struct param _params[] = {
-		{0, 1, ESTR("invalid number of feature args")},
+		{0, 3, "invalid number of feature args"},
+		{1, 50, "pg_init_retries must be between 1 and 50"},
 	};
 
 	r = read_param(_params, shift(as), &argc, &ti->error);
@@ -667,12 +788,28 @@ static int parse_features(struct arg_set *as, struct multipath *m,
 	if (!argc)
 		return 0;
 
-	if (!strnicmp(shift(as), MESG_STR("queue_if_no_path")))
-		return queue_if_no_path(m, 1);
-	else {
+	do {
+		param_name = shift(as);
+		argc--;
+
+		if (!strnicmp(param_name, MESG_STR("queue_if_no_path"))) {
+			r = queue_if_no_path(m, 1, 0);
+			continue;
+		}
+
+		if (!strnicmp(param_name, MESG_STR("pg_init_retries")) &&
+		    (argc >= 1)) {
+			r = read_param(_params + 1, shift(as),
+				       &m->pg_init_retries, &ti->error);
+			argc--;
+			continue;
+		}
+
 		ti->error = "Unrecognised multipath feature request";
-		return -EINVAL;
-	}
+		r = -EINVAL;
+	} while (argc && !r);
+
+	return r;
 }
 
 static int multipath_ctr(struct dm_target *ti, unsigned int argc,
@@ -680,8 +817,8 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 {
 	/* target parameters */
 	static struct param _params[] = {
-		{1, 1024, ESTR("invalid number of priority groups")},
-		{1, 1024, ESTR("invalid initial priority group number")},
+		{1, 1024, "invalid number of priority groups"},
+		{1, 1024, "invalid initial priority group number"},
 	};
 
 	int r;
@@ -693,17 +830,17 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	as.argc = argc;
 	as.argv = argv;
 
-	m = alloc_multipath();
+	m = alloc_multipath(ti);
 	if (!m) {
-		ti->error = ESTR("can't allocate multipath");
+		ti->error = "can't allocate multipath";
 		return -EINVAL;
 	}
 
-	r = parse_features(&as, m, ti);
+	r = parse_features(&as, m);
 	if (r)
 		goto bad;
 
-	r = parse_hw_handler(&as, m, ti);
+	r = parse_hw_handler(&as, m);
 	if (r)
 		goto bad;
 
@@ -719,9 +856,9 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	while (as.argc) {
 		struct priority_group *pg;
 
-		pg = parse_priority_group(&as, m, ti);
-		if (!pg) {
-			r = -EINVAL;
+		pg = parse_priority_group(&as, m);
+		if (IS_ERR(pg)) {
+			r = PTR_ERR(pg);
 			goto bad;
 		}
 
@@ -734,13 +871,12 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	}
 
 	if (pg_count != m->nr_priority_groups) {
-		ti->error = ESTR("priority group count mismatch");
+		ti->error = "priority group count mismatch";
 		r = -EINVAL;
 		goto bad;
 	}
 
-	ti->private = m;
-	m->ti = ti;
+	ti->num_flush_requests = 1;
 
 	return 0;
 
@@ -752,29 +888,33 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 static void multipath_dtr(struct dm_target *ti)
 {
 	struct multipath *m = (struct multipath *) ti->private;
+
+	flush_workqueue(kmpath_handlerd);
+	flush_workqueue(kmultipathd);
+	flush_scheduled_work();
 	free_multipath(m);
 }
 
 /*
- * Map bios, recording original fields for later in case we have to resubmit
+ * Map cloned requests
  */
-static int multipath_map(struct dm_target *ti, struct bio *bio,
+static int multipath_map(struct dm_target *ti, struct request *clone,
 			 union map_info *map_context)
 {
 	int r;
-	struct mpath_io *mpio;
+	struct dm_mpath_io *mpio;
 	struct multipath *m = (struct multipath *) ti->private;
 
-	if (bio_barrier(bio))
-		return -EOPNOTSUPP;
-
-	mpio = mempool_alloc(m->mpio_pool, GFP_NOIO);
-	dm_bio_record(&mpio->details, bio);
+	mpio = mempool_alloc(m->mpio_pool, GFP_ATOMIC);
+	if (!mpio)
+		/* ENOMEM, requeue */
+		return DM_MAPIO_REQUEUE;
+	memset(mpio, 0, sizeof(*mpio));
 
 	map_context->ptr = mpio;
-	bio->bi_rw |= (1 << BIO_RW_FAILFAST);
-	r = map_io(m, bio, mpio, 0);
-	if (r < 0)
+	clone->cmd_flags |= REQ_FAILFAST_TRANSPORT;
+	r = map_io(m, clone, mpio, 0);
+	if (r < 0 || r == DM_MAPIO_REQUEUE)
 		mempool_free(mpio, m->mpio_pool);
 
 	return r;
@@ -790,13 +930,13 @@ static int fail_path(struct pgpath *pgpath)
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (!pgpath->path.is_active)
+	if (!pgpath->is_active)
 		goto out;
 
-	DMWARN("dm-multipath: Failing path %s.", pgpath->path.dev->name);
+	DMWARN("Failing path %s.", pgpath->path.dev->name);
 
 	pgpath->pg->ps.type->fail_path(&pgpath->pg->ps, &pgpath->path);
-	pgpath->path.is_active = 0;
+	pgpath->is_active = 0;
 	pgpath->fail_count++;
 
 	m->nr_valid_paths--;
@@ -804,7 +944,11 @@ static int fail_path(struct pgpath *pgpath)
 	if (pgpath == m->current_pgpath)
 		m->current_pgpath = NULL;
 
-	queue_work(kmultipathd, &m->trigger_event);
+	dm_path_uevent(DM_UEVENT_PATH_FAILED, m->ti,
+		      pgpath->path.dev->name, m->nr_valid_paths);
+
+	schedule_work(&m->trigger_event);
+	queue_work(kmultipathd, &pgpath->deactivate_path);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -823,10 +967,10 @@ static int reinstate_path(struct pgpath *pgpath)
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (pgpath->path.is_active)
+	if (pgpath->is_active)
 		goto out;
 
-	if (!pgpath->pg->ps.type) {
+	if (!pgpath->pg->ps.type->reinstate_path) {
 		DMWARN("Reinstate path not supported by path selector %s",
 		       pgpath->pg->ps.type->name);
 		r = -EINVAL;
@@ -837,13 +981,20 @@ static int reinstate_path(struct pgpath *pgpath)
 	if (r)
 		goto out;
 
-	pgpath->path.is_active = 1;
+	pgpath->is_active = 1;
 
-	m->current_pgpath = NULL;
-	if (!m->nr_valid_paths++)
+	if (!m->nr_valid_paths++ && m->queue_size) {
+		m->current_pgpath = NULL;
 		queue_work(kmultipathd, &m->process_queued_ios);
+	} else if (m->hw_handler_name && (m->current_pg == pgpath->pg)) {
+		if (queue_work(kmpath_handlerd, &pgpath->activate_path))
+			m->pg_init_in_progress++;
+	}
 
-	queue_work(kmultipathd, &m->trigger_event);
+	dm_path_uevent(DM_UEVENT_PATH_REINSTATED, m->ti,
+		      pgpath->path.dev->name, m->nr_valid_paths);
+
+	schedule_work(&m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -887,7 +1038,7 @@ static void bypass_pg(struct multipath *m, struct priority_group *pg,
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 }
 
 /*
@@ -917,7 +1068,7 @@ static int switch_pg_num(struct multipath *m, const char *pgstr)
 	}
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 	return 0;
 }
 
@@ -946,138 +1097,180 @@ static int bypass_pg_num(struct multipath *m, const char *pgstr, int bypassed)
 }
 
 /*
- * pg_init must call this when it has completed its initialisation
+ * Should we retry pg_init immediately?
  */
-void dm_pg_init_complete(struct path *path, unsigned err_flags)
+static int pg_init_limit_reached(struct multipath *m, struct pgpath *pgpath)
+{
+	unsigned long flags;
+	int limit_reached = 0;
+
+	spin_lock_irqsave(&m->lock, flags);
+
+	if (m->pg_init_count <= m->pg_init_retries)
+		m->pg_init_required = 1;
+	else
+		limit_reached = 1;
+
+	spin_unlock_irqrestore(&m->lock, flags);
+
+	return limit_reached;
+}
+
+static void pg_init_done(void *data, int errors)
 {
+	struct dm_path *path = data;
 	struct pgpath *pgpath = path_to_pgpath(path);
 	struct priority_group *pg = pgpath->pg;
 	struct multipath *m = pg->m;
 	unsigned long flags;
 
-	/* We insist on failing the path if the PG is already bypassed. */
-	if (err_flags && pg->bypassed)
-		err_flags |= MP_FAIL_PATH;
-
-	if (err_flags & MP_FAIL_PATH)
+	/* device or driver problems */
+	switch (errors) {
+	case SCSI_DH_OK:
+		break;
+	case SCSI_DH_NOSYS:
+		if (!m->hw_handler_name) {
+			errors = 0;
+			break;
+		}
+		DMERR("Cannot failover device because scsi_dh_%s was not "
+		      "loaded.", m->hw_handler_name);
+		/*
+		 * Fail path for now, so we do not ping pong
+		 */
 		fail_path(pgpath);
-
-	if (err_flags & MP_BYPASS_PG)
+		break;
+	case SCSI_DH_DEV_TEMP_BUSY:
+		/*
+		 * Probably doing something like FW upgrade on the
+		 * controller so try the other pg.
+		 */
 		bypass_pg(m, pg, 1);
+		break;
+	/* TODO: For SCSI_DH_RETRY we should wait a couple seconds */
+	case SCSI_DH_RETRY:
+	case SCSI_DH_IMM_RETRY:
+	case SCSI_DH_RES_TEMP_UNAVAIL:
+		if (pg_init_limit_reached(m, pgpath))
+			fail_path(pgpath);
+		errors = 0;
+		break;
+	default:
+		/*
+		 * We probably do not want to fail the path for a device
+		 * error, but this is what the old dm did. In future
+		 * patches we can do more advanced handling.
+		 */
+		fail_path(pgpath);
+	}
 
 	spin_lock_irqsave(&m->lock, flags);
-	if (!err_flags)
+	if (errors) {
+		if (pgpath == m->current_pgpath) {
+			DMERR("Could not failover device. Error %d.", errors);
+			m->current_pgpath = NULL;
+			m->current_pg = NULL;
+		}
+	} else if (!m->pg_init_required) {
 		m->queue_io = 0;
-	else {
-		m->current_pgpath = NULL;
-		m->current_pg = NULL;
+		pg->bypassed = 0;
 	}
-	queue_work(kmultipathd, &m->process_queued_ios);
+
+	m->pg_init_in_progress--;
+	if (!m->pg_init_in_progress)
+		queue_work(kmultipathd, &m->process_queued_ios);
 	spin_unlock_irqrestore(&m->lock, flags);
 }
 
+static void activate_path(struct work_struct *work)
+{
+	struct pgpath *pgpath =
+		container_of(work, struct pgpath, activate_path);
+
+	scsi_dh_activate(bdev_get_queue(pgpath->path.dev->bdev),
+				pg_init_done, &pgpath->path);
+}
+
 /*
  * end_io handling
  */
-static int do_end_io(struct multipath *m, struct bio *bio,
-		     int error, struct mpath_io *mpio)
+static int do_end_io(struct multipath *m, struct request *clone,
+		     int error, struct dm_mpath_io *mpio)
 {
-	struct hw_handler *hwh = &m->hw_handler;
-	unsigned err_flags = MP_FAIL_PATH;	/* Default behavior */
+	/*
+	 * We don't queue any clone request inside the multipath target
+	 * during end I/O handling, since those clone requests don't have
+	 * bio clones.  If we queue them inside the multipath target,
+	 * we need to make bio clones, that requires memory allocation.
+	 * (See drivers/md/dm.c:end_clone_bio() about why the clone requests
+	 *  don't have bio clones.)
+	 * Instead of queueing the clone request here, we queue the original
+	 * request into dm core, which will remake a clone request and
+	 * clone bios for it and resubmit it later.
+	 */
+	int r = DM_ENDIO_REQUEUE;
+	unsigned long flags;
 
-	if (!error)
+	if (!error && !clone->errors)
 		return 0;	/* I/O complete */
 
-	if ((error == -EWOULDBLOCK) && bio_rw_ahead(bio))
-		return error;
-
 	if (error == -EOPNOTSUPP)
 		return error;
 
-	spin_lock(&m->lock);
-	if (!m->nr_valid_paths) {
-		if (!m->queue_if_no_path || m->suspended) {
-			spin_unlock(&m->lock);
-			return -EIO;
-		} else {
-			spin_unlock(&m->lock);
-			goto requeue;
-		}
-	}
-	spin_unlock(&m->lock);
-
-	if (hwh->type && hwh->type->error)
-		err_flags = hwh->type->error(hwh, bio);
-
-	if (mpio->pgpath) {
-		if (err_flags & MP_FAIL_PATH)
-			fail_path(mpio->pgpath);
-
-		if (err_flags & MP_BYPASS_PG)
-			bypass_pg(m, mpio->pgpath->pg, 1);
-	}
+	if (mpio->pgpath)
+		fail_path(mpio->pgpath);
 
-	if (err_flags & MP_ERROR_IO)
-		return -EIO;
-
-      requeue:
-	dm_bio_restore(&mpio->details, bio);
-
-	/* queue for the daemon to resubmit or fail */
-	spin_lock(&m->lock);
-	bio_list_add(&m->queued_ios, bio);
-	m->queue_size++;
-	if (!m->queue_io)
-		queue_work(kmultipathd, &m->process_queued_ios);
-	spin_unlock(&m->lock);
+	spin_lock_irqsave(&m->lock, flags);
+	if (!m->nr_valid_paths && !m->queue_if_no_path && !__must_push_back(m))
+		r = -EIO;
+	spin_unlock_irqrestore(&m->lock, flags);
 
-	return 1;	/* io not complete */
+	return r;
 }
 
-static int multipath_end_io(struct dm_target *ti, struct bio *bio,
+static int multipath_end_io(struct dm_target *ti, struct request *clone,
 			    int error, union map_info *map_context)
 {
-	struct multipath *m = (struct multipath *) ti->private;
-	struct mpath_io *mpio = (struct mpath_io *) map_context->ptr;
+	struct multipath *m = ti->private;
+	struct dm_mpath_io *mpio = map_context->ptr;
 	struct pgpath *pgpath = mpio->pgpath;
 	struct path_selector *ps;
 	int r;
 
-	r  = do_end_io(m, bio, error, mpio);
+	r  = do_end_io(m, clone, error, mpio);
 	if (pgpath) {
 		ps = &pgpath->pg->ps;
 		if (ps->type->end_io)
-			ps->type->end_io(ps, &pgpath->path);
+			ps->type->end_io(ps, &pgpath->path, mpio->nr_bytes);
 	}
-	if (r <= 0)
-		mempool_free(mpio, m->mpio_pool);
+	mempool_free(mpio, m->mpio_pool);
 
 	return r;
 }
 
 /*
  * Suspend can't complete until all the I/O is processed so if
- * the last path failed we will now error any queued I/O.
+ * the last path fails we must error any remaining I/O.
+ * Note that if the freeze_bdev fails while suspending, the
+ * queue_if_no_path state is lost - userspace should reset it.
  */
 static void multipath_presuspend(struct dm_target *ti)
 {
 	struct multipath *m = (struct multipath *) ti->private;
-	unsigned long flags;
 
-	spin_lock_irqsave(&m->lock, flags);
-	m->suspended = 1;
-	if (m->queue_if_no_path)
-		queue_work(kmultipathd, &m->process_queued_ios);
-	spin_unlock_irqrestore(&m->lock, flags);
+	queue_if_no_path(m, 0, 1);
 }
 
+/*
+ * Restore the queue_if_no_path setting.
+ */
 static void multipath_resume(struct dm_target *ti)
 {
 	struct multipath *m = (struct multipath *) ti->private;
 	unsigned long flags;
 
 	spin_lock_irqsave(&m->lock, flags);
-	m->suspended = 0;
+	m->queue_if_no_path = m->saved_queue_if_no_path;
 	spin_unlock_irqrestore(&m->lock, flags);
 }
 
@@ -1103,7 +1296,6 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 	int sz = 0;
 	unsigned long flags;
 	struct multipath *m = (struct multipath *) ti->private;
-	struct hw_handler *hwh = &m->hw_handler;
 	struct priority_group *pg;
 	struct pgpath *p;
 	unsigned pg_num;
@@ -1113,18 +1305,20 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 	/* Features */
 	if (type == STATUSTYPE_INFO)
-		DMEMIT("1 %u ", m->queue_size);
-	else if (m->queue_if_no_path)
-		DMEMIT("1 queue_if_no_path ");
-	else
-		DMEMIT("0 ");
+		DMEMIT("2 %u %u ", m->queue_size, m->pg_init_count);
+	else {
+		DMEMIT("%u ", m->queue_if_no_path +
+			      (m->pg_init_retries > 0) * 2);
+		if (m->queue_if_no_path)
+			DMEMIT("queue_if_no_path ");
+		if (m->pg_init_retries)
+			DMEMIT("pg_init_retries %u ", m->pg_init_retries);
+	}
 
-	if (hwh->type && hwh->type->status)
-		sz += hwh->type->status(hwh, type, result + sz, maxlen - sz);
-	else if (!hwh->type || type == STATUSTYPE_INFO)
+	if (!m->hw_handler_name || type == STATUSTYPE_INFO)
 		DMEMIT("0 ");
 	else
-		DMEMIT("1 %s ", hwh->type->name);
+		DMEMIT("1 %s ", m->hw_handler_name);
 
 	DMEMIT("%u ", m->nr_priority_groups);
 
@@ -1161,7 +1355,7 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 			list_for_each_entry(p, &pg->pgpaths, list) {
 				DMEMIT("%s %s %u ", p->path.dev->name,
-				       p->path.is_active ? "A" : "F",
+				       p->is_active ? "A" : "F",
 				       p->fail_count);
 				if (pg->ps.type->status)
 					sz += pg->ps.type->status(&pg->ps,
@@ -1210,9 +1404,9 @@ static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 
 	if (argc == 1) {
 		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path")))
-			return queue_if_no_path(m, 1);
+			return queue_if_no_path(m, 1, 0);
 		else if (!strnicmp(argv[0], MESG_STR("fail_if_no_path")))
-			return queue_if_no_path(m, 0);
+			return queue_if_no_path(m, 0, 0);
 	}
 
 	if (argc != 2)
@@ -1234,7 +1428,7 @@ static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 	r = dm_get_device(ti, argv[1], ti->begin, ti->len,
 			  dm_table_get_mode(ti->table), &dev);
 	if (r) {
-		DMWARN("dm-multipath message: error getting device %s",
+		DMWARN("message: error getting device %s",
 		       argv[1]);
 		return -EINVAL;
 	}
@@ -1250,21 +1444,142 @@ error:
 	return -EINVAL;
 }
 
+static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
+			   unsigned long arg)
+{
+	struct multipath *m = (struct multipath *) ti->private;
+	struct block_device *bdev = NULL;
+	fmode_t mode = 0;
+	unsigned long flags;
+	int r = 0;
+
+	spin_lock_irqsave(&m->lock, flags);
+
+	if (!m->current_pgpath)
+		__choose_pgpath(m, 0);
+
+	if (m->current_pgpath) {
+		bdev = m->current_pgpath->path.dev->bdev;
+		mode = m->current_pgpath->path.dev->mode;
+	}
+
+	if (m->queue_io)
+		r = -EAGAIN;
+	else if (!bdev)
+		r = -EIO;
+
+	spin_unlock_irqrestore(&m->lock, flags);
+
+	return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);
+}
+
+static int multipath_iterate_devices(struct dm_target *ti,
+				     iterate_devices_callout_fn fn, void *data)
+{
+	struct multipath *m = ti->private;
+	struct priority_group *pg;
+	struct pgpath *p;
+	int ret = 0;
+
+	list_for_each_entry(pg, &m->priority_groups, list) {
+		list_for_each_entry(p, &pg->pgpaths, list) {
+			ret = fn(ti, p->path.dev, ti->begin, ti->len, data);
+			if (ret)
+				goto out;
+		}
+	}
+
+out:
+	return ret;
+}
+
+static int __pgpath_busy(struct pgpath *pgpath)
+{
+	struct request_queue *q = bdev_get_queue(pgpath->path.dev->bdev);
+
+	return dm_underlying_device_busy(q);
+}
+
+/*
+ * We return "busy", only when we can map I/Os but underlying devices
+ * are busy (so even if we map I/Os now, the I/Os will wait on
+ * the underlying queue).
+ * In other words, if we want to kill I/Os or queue them inside us
+ * due to map unavailability, we don't return "busy".  Otherwise,
+ * dm core won't give us the I/Os and we can't do what we want.
+ */
+static int multipath_busy(struct dm_target *ti)
+{
+	int busy = 0, has_active = 0;
+	struct multipath *m = ti->private;
+	struct priority_group *pg;
+	struct pgpath *pgpath;
+	unsigned long flags;
+
+	spin_lock_irqsave(&m->lock, flags);
+
+	/* Guess which priority_group will be used at next mapping time */
+	if (unlikely(!m->current_pgpath && m->next_pg))
+		pg = m->next_pg;
+	else if (likely(m->current_pg))
+		pg = m->current_pg;
+	else
+		/*
+		 * We don't know which pg will be used at next mapping time.
+		 * We don't call __choose_pgpath() here to avoid to trigger
+		 * pg_init just by busy checking.
+		 * So we don't know whether underlying devices we will be using
+		 * at next mapping time are busy or not. Just try mapping.
+		 */
+		goto out;
+
+	/*
+	 * If there is one non-busy active path at least, the path selector
+	 * will be able to select it. So we consider such a pg as not busy.
+	 */
+	busy = 1;
+	list_for_each_entry(pgpath, &pg->pgpaths, list)
+		if (pgpath->is_active) {
+			has_active = 1;
+
+			if (!__pgpath_busy(pgpath)) {
+				busy = 0;
+				break;
+			}
+		}
+
+	if (!has_active)
+		/*
+		 * No active path in this pg, so this pg won't be used and
+		 * the current_pg will be changed at next mapping time.
+		 * We need to try mapping to determine it.
+		 */
+		busy = 0;
+
+out:
+	spin_unlock_irqrestore(&m->lock, flags);
+
+	return busy;
+}
+
 /*-----------------------------------------------------------------
  * Module setup
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 0, 4},
+	.version = {1, 1, 0},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,
-	.map = multipath_map,
-	.end_io = multipath_end_io,
+	.map_rq = multipath_map,
+	.rq_end_io = multipath_end_io,
 	.presuspend = multipath_presuspend,
 	.resume = multipath_resume,
 	.status = multipath_status,
 	.message = multipath_message,
+	.ioctl  = multipath_ioctl,
+	.iterate_devices = multipath_iterate_devices,
+	.busy = multipath_busy,
 };
 
 static int __init dm_multipath_init(void)
@@ -1272,28 +1587,41 @@ static int __init dm_multipath_init(void)
 	int r;
 
 	/* allocate a slab for the dm_ios */
-	_mpio_cache = kmem_cache_create("dm_mpath", sizeof(struct mpath_io),
-					0, 0, NULL, NULL);
+	_mpio_cache = KMEM_CACHE(dm_mpath_io, 0);
 	if (!_mpio_cache)
 		return -ENOMEM;
 
 	r = dm_register_target(&multipath_target);
 	if (r < 0) {
-		DMERR("%s: register failed %d", multipath_target.name, r);
+		DMERR("register failed %d", r);
 		kmem_cache_destroy(_mpio_cache);
 		return -EINVAL;
 	}
 
 	kmultipathd = create_workqueue("kmpathd");
 	if (!kmultipathd) {
-		DMERR("%s: failed to create workqueue kmpathd",
-				multipath_target.name);
+		DMERR("failed to create workqueue kmpathd");
+		dm_unregister_target(&multipath_target);
+		kmem_cache_destroy(_mpio_cache);
+		return -ENOMEM;
+	}
+
+	/*
+	 * A separate workqueue is used to handle the device handlers
+	 * to avoid overloading existing workqueue. Overloading the
+	 * old workqueue would also create a bottleneck in the
+	 * path of the storage hardware device activation.
+	 */
+	kmpath_handlerd = create_singlethread_workqueue("kmpath_handlerd");
+	if (!kmpath_handlerd) {
+		DMERR("failed to create workqueue kmpath_handlerd");
+		destroy_workqueue(kmultipathd);
 		dm_unregister_target(&multipath_target);
 		kmem_cache_destroy(_mpio_cache);
 		return -ENOMEM;
 	}
 
-	DMINFO("dm-multipath version %u.%u.%u loaded",
+	DMINFO("version %u.%u.%u loaded",
 	       multipath_target.version[0], multipath_target.version[1],
 	       multipath_target.version[2]);
 
@@ -1302,19 +1630,13 @@ static int __init dm_multipath_init(void)
 
 static void __exit dm_multipath_exit(void)
 {
-	int r;
-
+	destroy_workqueue(kmpath_handlerd);
 	destroy_workqueue(kmultipathd);
 
-	r = dm_unregister_target(&multipath_target);
-	if (r < 0)
-		DMERR("%s: target unregister failed %d",
-		      multipath_target.name, r);
+	dm_unregister_target(&multipath_target);
 	kmem_cache_destroy(_mpio_cache);
 }
 
-EXPORT_SYMBOL_GPL(dm_pg_init_complete);
-
 module_init(dm_multipath_init);
 module_exit(dm_multipath_exit);