Merge branch 'quilt/device-mapper'

16 files changed, 859 insertions, 553 deletions

diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index 7dbe652efb5a..2b7907b6dd09 100644
--- a/drivers/md/dm-exception-store.c
+++ b/drivers/md/dm-exception-store.c
@@ -172,7 +172,8 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 	}
 
 	/* Validate the chunk size against the device block size */
-	if (chunk_size % (bdev_logical_block_size(store->cow->bdev) >> 9)) {
+	if (chunk_size %
+	    (bdev_logical_block_size(dm_snap_cow(store->snap)->bdev) >> 9)) {
 		*error = "Chunk size is not a multiple of device blocksize";
 		return -EINVAL;
 	}
@@ -190,6 +191,7 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 }
 
 int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
 			      unsigned *args_used,
 			      struct dm_exception_store **store)
 {
@@ -198,7 +200,7 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 	struct dm_exception_store *tmp_store;
 	char persistent;
 
-	if (argc < 3) {
+	if (argc < 2) {
 		ti->error = "Insufficient exception store arguments";
 		return -EINVAL;
 	}
@@ -209,14 +211,15 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 		return -ENOMEM;
 	}
 
-	persistent = toupper(*argv[1]);
+	persistent = toupper(*argv[0]);
 	if (persistent == 'P')
 		type = get_type("P");
 	else if (persistent == 'N')
 		type = get_type("N");
 	else {
 		ti->error = "Persistent flag is not P or N";
-		return -EINVAL;
+		r = -EINVAL;
+		goto bad_type;
 	}
 
 	if (!type) {
@@ -226,32 +229,23 @@ int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
 	}
 
 	tmp_store->type = type;
-	tmp_store->ti = ti;
-
-	r = dm_get_device(ti, argv[0], 0, 0,
-			  FMODE_READ | FMODE_WRITE, &tmp_store->cow);
-	if (r) {
-		ti->error = "Cannot get COW device";
-		goto bad_cow;
-	}
+	tmp_store->snap = snap;
 
-	r = set_chunk_size(tmp_store, argv[2], &ti->error);
+	r = set_chunk_size(tmp_store, argv[1], &ti->error);
 	if (r)
-		goto bad_ctr;
+		goto bad;
 
 	r = type->ctr(tmp_store, 0, NULL);
 	if (r) {
 		ti->error = "Exception store type constructor failed";
-		goto bad_ctr;
+		goto bad;
 	}
 
-	*args_used = 3;
+	*args_used = 2;
 	*store = tmp_store;
 	return 0;
 
-bad_ctr:
-	dm_put_device(ti, tmp_store->cow);
-bad_cow:
+bad:
 	put_type(type);
 bad_type:
 	kfree(tmp_store);
@@ -262,7 +256,6 @@ EXPORT_SYMBOL(dm_exception_store_create);
 void dm_exception_store_destroy(struct dm_exception_store *store)
 {
 	store->type->dtr(store);
-	dm_put_device(store->ti, store->cow);
 	put_type(store->type);
 	kfree(store);
 }
diff --git a/drivers/md/dm-exception-store.h b/drivers/md/dm-exception-store.h
index 8a223a48802c..bb8874653de1 100644
--- a/drivers/md/dm-exception-store.h
+++ b/drivers/md/dm-exception-store.h
@@ -26,7 +26,7 @@ typedef sector_t chunk_t;
  * of chunks that follow contiguously.  Remaining bits hold the number of the
  * chunk within the device.
  */
-struct dm_snap_exception {
+struct dm_exception {
 	struct list_head hash_list;
 
 	chunk_t old_chunk;
@@ -64,13 +64,13 @@ struct dm_exception_store_type {
 	 * Find somewhere to store the next exception.
 	 */
 	int (*prepare_exception) (struct dm_exception_store *store,
-				  struct dm_snap_exception *e);
+				  struct dm_exception *e);
 
 	/*
 	 * Update the metadata with this exception.
 	 */
 	void (*commit_exception) (struct dm_exception_store *store,
-				  struct dm_snap_exception *e,
+				  struct dm_exception *e,
 				  void (*callback) (void *, int success),
 				  void *callback_context);
 
@@ -86,19 +86,19 @@ struct dm_exception_store_type {
 	/*
 	 * Return how full the snapshot is.
 	 */
-	void (*fraction_full) (struct dm_exception_store *store,
-			       sector_t *numerator,
-			       sector_t *denominator);
+	void (*usage) (struct dm_exception_store *store,
+		       sector_t *total_sectors, sector_t *sectors_allocated,
+		       sector_t *metadata_sectors);
 
 	/* For internal device-mapper use only. */
 	struct list_head list;
 };
 
+struct dm_snapshot;
+
 struct dm_exception_store {
 	struct dm_exception_store_type *type;
-	struct dm_target *ti;
-
-	struct dm_dev *cow;
+	struct dm_snapshot *snap;
 
 	/* Size of data blocks saved - must be a power of 2 */
 	unsigned chunk_size;
@@ -109,6 +109,11 @@ struct dm_exception_store {
 };
 
 /*
+ * Obtain the cow device used by a given snapshot.
+ */
+struct dm_dev *dm_snap_cow(struct dm_snapshot *snap);
+
+/*
  * Funtions to manipulate consecutive chunks
  */
 #  if defined(CONFIG_LBDAF) || (BITS_PER_LONG == 64)
@@ -120,12 +125,12 @@ static inline chunk_t dm_chunk_number(chunk_t chunk)
 	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
 }
 
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
 {
 	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
 }
 
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
 {
 	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
 
@@ -140,12 +145,12 @@ static inline chunk_t dm_chunk_number(chunk_t chunk)
 	return chunk;
 }
 
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
 {
 	return 0;
 }
 
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
 {
 }
 
@@ -162,7 +167,7 @@ static inline sector_t get_dev_size(struct block_device *bdev)
 static inline chunk_t sector_to_chunk(struct dm_exception_store *store,
 				      sector_t sector)
 {
-	return (sector & ~store->chunk_mask) >> store->chunk_shift;
+	return sector >> store->chunk_shift;
 }
 
 int dm_exception_store_type_register(struct dm_exception_store_type *type);
@@ -173,6 +178,7 @@ int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
 				      char **error);
 
 int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
 			      unsigned *args_used,
 			      struct dm_exception_store **store);
 void dm_exception_store_destroy(struct dm_exception_store *store);
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index 3a2e6a2f8bdd..10f457ca6af2 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -5,6 +5,8 @@
  * This file is released under the GPL.
  */
 
+#include "dm.h"
+
 #include <linux/device-mapper.h>
 
 #include <linux/bio.h>
@@ -14,12 +16,19 @@
 #include <linux/slab.h>
 #include <linux/dm-io.h>
 
+#define DM_MSG_PREFIX "io"
+
+#define DM_IO_MAX_REGIONS	BITS_PER_LONG
+
 struct dm_io_client {
 	mempool_t *pool;
 	struct bio_set *bios;
 };
 
-/* FIXME: can we shrink this ? */
+/*
+ * Aligning 'struct io' reduces the number of bits required to store
+ * its address.  Refer to store_io_and_region_in_bio() below.
+ */
 struct io {
 	unsigned long error_bits;
 	unsigned long eopnotsupp_bits;
@@ -28,7 +37,9 @@ struct io {
 	struct dm_io_client *client;
 	io_notify_fn callback;
 	void *context;
-};
+} __attribute__((aligned(DM_IO_MAX_REGIONS)));
+
+static struct kmem_cache *_dm_io_cache;
 
 /*
  * io contexts are only dynamically allocated for asynchronous
@@ -53,7 +64,7 @@ struct dm_io_client *dm_io_client_create(unsigned num_pages)
 	if (!client)
 		return ERR_PTR(-ENOMEM);
 
-	client->pool = mempool_create_kmalloc_pool(ios, sizeof(struct io));
+	client->pool = mempool_create_slab_pool(ios, _dm_io_cache);
 	if (!client->pool)
 		goto bad;
 
@@ -88,18 +99,29 @@ EXPORT_SYMBOL(dm_io_client_destroy);
 
 /*-----------------------------------------------------------------
  * We need to keep track of which region a bio is doing io for.
- * In order to save a memory allocation we store this the last
- * bvec which we know is unused (blech).
- * XXX This is ugly and can OOPS with some configs... find another way.
+ * To avoid a memory allocation to store just 5 or 6 bits, we
+ * ensure the 'struct io' pointer is aligned so enough low bits are
+ * always zero and then combine it with the region number directly in
+ * bi_private.
  *---------------------------------------------------------------*/
-static inline void bio_set_region(struct bio *bio, unsigned region)
+static void store_io_and_region_in_bio(struct bio *bio, struct io *io,
+				       unsigned region)
 {
-	bio->bi_io_vec[bio->bi_max_vecs].bv_len = region;
+	if (unlikely(!IS_ALIGNED((unsigned long)io, DM_IO_MAX_REGIONS))) {
+		DMCRIT("Unaligned struct io pointer %p", io);
+		BUG();
+	}
+
+	bio->bi_private = (void *)((unsigned long)io | region);
 }
 
-static inline unsigned bio_get_region(struct bio *bio)
+static void retrieve_io_and_region_from_bio(struct bio *bio, struct io **io,
+				       unsigned *region)
 {
-	return bio->bi_io_vec[bio->bi_max_vecs].bv_len;
+	unsigned long val = (unsigned long)bio->bi_private;
+
+	*io = (void *)(val & -(unsigned long)DM_IO_MAX_REGIONS);
+	*region = val & (DM_IO_MAX_REGIONS - 1);
 }
 
 /*-----------------------------------------------------------------
@@ -140,10 +162,8 @@ static void endio(struct bio *bio, int error)
 	/*
 	 * The bio destructor in bio_put() may use the io object.
 	 */
-	io = bio->bi_private;
-	region = bio_get_region(bio);
+	retrieve_io_and_region_from_bio(bio, &io, &region);
 
-	bio->bi_max_vecs++;
 	bio_put(bio);
 
 	dec_count(io, region, error);
@@ -243,7 +263,10 @@ static void vm_dp_init(struct dpages *dp, void *data)
 
 static void dm_bio_destructor(struct bio *bio)
 {
-	struct io *io = bio->bi_private;
+	unsigned region;
+	struct io *io;
+
+	retrieve_io_and_region_from_bio(bio, &io, &region);
 
 	bio_free(bio, io->client->bios);
 }
@@ -286,26 +309,23 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 	unsigned num_bvecs;
 	sector_t remaining = where->count;
 
-	while (remaining) {
+	/*
+	 * where->count may be zero if rw holds a write barrier and we
+	 * need to send a zero-sized barrier.
+	 */
+	do {
 		/*
-		 * Allocate a suitably sized-bio: we add an extra
-		 * bvec for bio_get/set_region() and decrement bi_max_vecs
-		 * to hide it from bio_add_page().
+		 * Allocate a suitably sized-bio.
 		 */
 		num_bvecs = dm_sector_div_up(remaining,
 					     (PAGE_SIZE >> SECTOR_SHIFT));
-		num_bvecs = 1 + min_t(int, bio_get_nr_vecs(where->bdev),
-				      num_bvecs);
-		if (unlikely(num_bvecs > BIO_MAX_PAGES))
-			num_bvecs = BIO_MAX_PAGES;
+		num_bvecs = min_t(int, bio_get_nr_vecs(where->bdev), num_bvecs);
 		bio = bio_alloc_bioset(GFP_NOIO, num_bvecs, io->client->bios);
 		bio->bi_sector = where->sector + (where->count - remaining);
 		bio->bi_bdev = where->bdev;
 		bio->bi_end_io = endio;
-		bio->bi_private = io;
 		bio->bi_destructor = dm_bio_destructor;
-		bio->bi_max_vecs--;
-		bio_set_region(bio, region);
+		store_io_and_region_in_bio(bio, io, region);
 
 		/*
 		 * Try and add as many pages as possible.
@@ -323,7 +343,7 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 
 		atomic_inc(&io->count);
 		submit_bio(rw, bio);
-	}
+	} while (remaining);
 }
 
 static void dispatch_io(int rw, unsigned int num_regions,
@@ -333,6 +353,8 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	int i;
 	struct dpages old_pages = *dp;
 
+	BUG_ON(num_regions > DM_IO_MAX_REGIONS);
+
 	if (sync)
 		rw |= (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
 
@@ -342,7 +364,7 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	 */
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
-		if (where[i].count)
+		if (where[i].count || (rw & (1 << BIO_RW_BARRIER)))
 			do_region(rw, i, where + i, dp, io);
 	}
 
@@ -357,7 +379,14 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, int rw, struct dpages *dp,
 		   unsigned long *error_bits)
 {
-	struct io io;
+	/*
+	 * gcc <= 4.3 can't do the alignment for stack variables, so we must
+	 * align it on our own.
+	 * volatile prevents the optimizer from removing or reusing
+	 * "io_" field from the stack frame (allowed in ANSI C).
+	 */
+	volatile char io_[sizeof(struct io) + __alignof__(struct io) - 1];
+	struct io *io = (struct io *)PTR_ALIGN(&io_, __alignof__(struct io));
 
 	if (num_regions > 1 && (rw & RW_MASK) != WRITE) {
 		WARN_ON(1);
@@ -365,33 +394,33 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	}
 
 retry:
-	io.error_bits = 0;
-	io.eopnotsupp_bits = 0;
-	atomic_set(&io.count, 1); /* see dispatch_io() */
-	io.sleeper = current;
-	io.client = client;
+	io->error_bits = 0;
+	io->eopnotsupp_bits = 0;
+	atomic_set(&io->count, 1); /* see dispatch_io() */
+	io->sleeper = current;
+	io->client = client;
 
-	dispatch_io(rw, num_regions, where, dp, &io, 1);
+	dispatch_io(rw, num_regions, where, dp, io, 1);
 
 	while (1) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 
-		if (!atomic_read(&io.count))
+		if (!atomic_read(&io->count))
 			break;
 
 		io_schedule();
 	}
 	set_current_state(TASK_RUNNING);
 
-	if (io.eopnotsupp_bits && (rw & (1 << BIO_RW_BARRIER))) {
+	if (io->eopnotsupp_bits && (rw & (1 << BIO_RW_BARRIER))) {
 		rw &= ~(1 << BIO_RW_BARRIER);
 		goto retry;
 	}
 
 	if (error_bits)
-		*error_bits = io.error_bits;
+		*error_bits = io->error_bits;
 
-	return io.error_bits ? -EIO : 0;
+	return io->error_bits ? -EIO : 0;
 }
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
@@ -472,3 +501,18 @@ int dm_io(struct dm_io_request *io_req, unsigned num_regions,
 			&dp, io_req->notify.fn, io_req->notify.context);
 }
 EXPORT_SYMBOL(dm_io);
+
+int __init dm_io_init(void)
+{
+	_dm_io_cache = KMEM_CACHE(io, 0);
+	if (!_dm_io_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void dm_io_exit(void)
+{
+	kmem_cache_destroy(_dm_io_cache);
+	_dm_io_cache = NULL;
+}
diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
index a67942931582..63fd4de25bda 100644
--- a/drivers/md/dm-ioctl.c
+++ b/drivers/md/dm-ioctl.c
@@ -56,6 +56,11 @@ static void dm_hash_remove_all(int keep_open_devices);
  */
 static DECLARE_RWSEM(_hash_lock);
 
+/*
+ * Protects use of mdptr to obtain hash cell name and uuid from mapped device.
+ */
+static DEFINE_MUTEX(dm_hash_cells_mutex);
+
 static void init_buckets(struct list_head *buckets)
 {
 	unsigned int i;
@@ -206,7 +211,9 @@ static int dm_hash_insert(const char *name, const char *uuid, struct mapped_devi
 		list_add(&cell->uuid_list, _uuid_buckets + hash_str(uuid));
 	}
 	dm_get(md);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(md, cell);
+	mutex_unlock(&dm_hash_cells_mutex);
 	up_write(&_hash_lock);
 
 	return 0;
@@ -224,9 +231,11 @@ static void __hash_remove(struct hash_cell *hc)
 	/* remove from the dev hash */
 	list_del(&hc->uuid_list);
 	list_del(&hc->name_list);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(hc->md, NULL);
+	mutex_unlock(&dm_hash_cells_mutex);
 
-	table = dm_get_table(hc->md);
+	table = dm_get_live_table(hc->md);
 	if (table) {
 		dm_table_event(table);
 		dm_table_put(table);
@@ -321,13 +330,15 @@ static int dm_hash_rename(uint32_t cookie, const char *old, const char *new)
 	 */
 	list_del(&hc->name_list);
 	old_name = hc->name;
+	mutex_lock(&dm_hash_cells_mutex);
 	hc->name = new_name;
+	mutex_unlock(&dm_hash_cells_mutex);
 	list_add(&hc->name_list, _name_buckets + hash_str(new_name));
 
 	/*
 	 * Wake up any dm event waiters.
 	 */
-	table = dm_get_table(hc->md);
+	table = dm_get_live_table(hc->md);
 	if (table) {
 		dm_table_event(table);
 		dm_table_put(table);
@@ -512,8 +523,6 @@ static int list_versions(struct dm_ioctl *param, size_t param_size)
 	return 0;
 }
 
-
-
 static int check_name(const char *name)
 {
 	if (strchr(name, '/')) {
@@ -525,6 +534,40 @@ static int check_name(const char *name)
 }
 
 /*
+ * On successful return, the caller must not attempt to acquire
+ * _hash_lock without first calling dm_table_put, because dm_table_destroy
+ * waits for this dm_table_put and could be called under this lock.
+ */
+static struct dm_table *dm_get_inactive_table(struct mapped_device *md)
+{
+	struct hash_cell *hc;
+	struct dm_table *table = NULL;
+
+	down_read(&_hash_lock);
+	hc = dm_get_mdptr(md);
+	if (!hc || hc->md != md) {
+		DMWARN("device has been removed from the dev hash table.");
+		goto out;
+	}
+
+	table = hc->new_map;
+	if (table)
+		dm_table_get(table);
+
+out:
+	up_read(&_hash_lock);
+
+	return table;
+}
+
+static struct dm_table *dm_get_live_or_inactive_table(struct mapped_device *md,
+						      struct dm_ioctl *param)
+{
+	return (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) ?
+		dm_get_inactive_table(md) : dm_get_live_table(md);
+}
+
+/*
  * Fills in a dm_ioctl structure, ready for sending back to
  * userland.
  */
@@ -548,18 +591,30 @@ static int __dev_status(struct mapped_device *md, struct dm_ioctl *param)
 	 */
 	param->open_count = dm_open_count(md);
 
-	if (get_disk_ro(disk))
-		param->flags |= DM_READONLY_FLAG;
-
 	param->event_nr = dm_get_event_nr(md);
+	param->target_count = 0;
 
-	table = dm_get_table(md);
+	table = dm_get_live_table(md);
 	if (table) {
-		param->flags |= DM_ACTIVE_PRESENT_FLAG;
-		param->target_count = dm_table_get_num_targets(table);
+		if (!(param->flags & DM_QUERY_INACTIVE_TABLE_FLAG)) {
+			if (get_disk_ro(disk))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+		}
 		dm_table_put(table);
-	} else
-		param->target_count = 0;
+
+		param->flags |= DM_ACTIVE_PRESENT_FLAG;
+	}
+
+	if (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) {
+		table = dm_get_inactive_table(md);
+		if (table) {
+			if (!(dm_table_get_mode(table) & FMODE_WRITE))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+			dm_table_put(table);
+		}
+	}
 
 	return 0;
 }
@@ -634,9 +689,9 @@ static struct mapped_device *find_device(struct dm_ioctl *param)
 		 * Sneakily write in both the name and the uuid
 		 * while we have the cell.
 		 */
-		strncpy(param->name, hc->name, sizeof(param->name));
+		strlcpy(param->name, hc->name, sizeof(param->name));
 		if (hc->uuid)
-			strncpy(param->uuid, hc->uuid, sizeof(param->uuid)-1);
+			strlcpy(param->uuid, hc->uuid, sizeof(param->uuid));
 		else
 			param->uuid[0] = '\0';
 
@@ -800,7 +855,7 @@ static int do_resume(struct dm_ioctl *param)
 	unsigned suspend_flags = DM_SUSPEND_LOCKFS_FLAG;
 	struct hash_cell *hc;
 	struct mapped_device *md;
-	struct dm_table *new_map;
+	struct dm_table *new_map, *old_map = NULL;
 
 	down_write(&_hash_lock);
 
@@ -829,11 +884,11 @@ static int do_resume(struct dm_ioctl *param)
 		if (!dm_suspended(md))
 			dm_suspend(md, suspend_flags);
 
-		r = dm_swap_table(md, new_map);
-		if (r) {
+		old_map = dm_swap_table(md, new_map);
+		if (IS_ERR(old_map)) {
 			dm_table_destroy(new_map);
 			dm_put(md);
-			return r;
+			return PTR_ERR(old_map);
 		}
 
 		if (dm_table_get_mode(new_map) & FMODE_WRITE)
@@ -845,6 +900,8 @@ static int do_resume(struct dm_ioctl *param)
 	if (dm_suspended(md))
 		r = dm_resume(md);
 
+	if (old_map)
+		dm_table_destroy(old_map);
 
 	if (!r) {
 		dm_kobject_uevent(md, KOBJ_CHANGE, param->event_nr);
@@ -982,7 +1039,7 @@ static int dev_wait(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_status(table, param, param_size);
 		dm_table_put(table);
@@ -1215,7 +1272,7 @@ static int table_deps(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_deps(table, param, param_size);
 		dm_table_put(table);
@@ -1244,13 +1301,13 @@ static int table_status(struct dm_ioctl *param, size_t param_size)
 	if (r)
 		goto out;
 
-	table = dm_get_table(md);
+	table = dm_get_live_or_inactive_table(md, param);
 	if (table) {
 		retrieve_status(table, param, param_size);
 		dm_table_put(table);
 	}
 
- out:
+out:
 	dm_put(md);
 	return r;
 }
@@ -1288,10 +1345,15 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		goto out;
 	}
 
-	table = dm_get_table(md);
+	table = dm_get_live_table(md);
 	if (!table)
 		goto out_argv;
 
+	if (dm_deleting_md(md)) {
+		r = -ENXIO;
+		goto out_table;
+	}
+
 	ti = dm_table_find_target(table, tmsg->sector);
 	if (!dm_target_is_valid(ti)) {
 		DMWARN("Target message sector outside device.");
@@ -1303,6 +1365,7 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		r = -EINVAL;
 	}
 
+ out_table:
 	dm_table_put(table);
  out_argv:
 	kfree(argv);
@@ -1582,8 +1645,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 	if (!md)
 		return -ENXIO;
 
-	dm_get(md);
-	down_read(&_hash_lock);
+	mutex_lock(&dm_hash_cells_mutex);
 	hc = dm_get_mdptr(md);
 	if (!hc || hc->md != md) {
 		r = -ENXIO;
@@ -1596,8 +1658,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 		strcpy(uuid, hc->uuid ? : "");
 
 out:
-	up_read(&_hash_lock);
-	dm_put(md);
+	mutex_unlock(&dm_hash_cells_mutex);
 
 	return r;
 }
diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c
index 3e3fc06cb861..addf83475040 100644
--- a/drivers/md/dm-kcopyd.c
+++ b/drivers/md/dm-kcopyd.c
@@ -450,7 +450,10 @@ static void dispatch_job(struct kcopyd_job *job)
 {
 	struct dm_kcopyd_client *kc = job->kc;
 	atomic_inc(&kc->nr_jobs);
-	push(&kc->pages_jobs, job);
+	if (unlikely(!job->source.count))
+		push(&kc->complete_jobs, job);
+	else
+		push(&kc->pages_jobs, job);
 	wake(kc);
 }
 
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index dce971dbdfa3..45d9bf14cc48 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -93,6 +93,10 @@ struct multipath {
 	 * can resubmit bios on error.
 	 */
 	mempool_t *mpio_pool;
+
+	struct mutex work_mutex;
+
+	unsigned suspended;	/* Don't create new I/O internally when set. */
 };
 
 /*
@@ -198,6 +202,7 @@ static struct multipath *alloc_multipath(struct dm_target *ti)
 		m->queue_io = 1;
 		INIT_WORK(&m->process_queued_ios, process_queued_ios);
 		INIT_WORK(&m->trigger_event, trigger_event);
+		mutex_init(&m->work_mutex);
 		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
 		if (!m->mpio_pool) {
 			kfree(m);
@@ -885,13 +890,18 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	return r;
 }
 
-static void multipath_dtr(struct dm_target *ti)
+static void flush_multipath_work(void)
 {
-	struct multipath *m = (struct multipath *) ti->private;
-
 	flush_workqueue(kmpath_handlerd);
 	flush_workqueue(kmultipathd);
 	flush_scheduled_work();
+}
+
+static void multipath_dtr(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	flush_multipath_work();
 	free_multipath(m);
 }
 
@@ -1261,6 +1271,16 @@ static void multipath_presuspend(struct dm_target *ti)
 	queue_if_no_path(m, 0, 1);
 }
 
+static void multipath_postsuspend(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	mutex_lock(&m->work_mutex);
+	m->suspended = 1;
+	flush_multipath_work();
+	mutex_unlock(&m->work_mutex);
+}
+
 /*
  * Restore the queue_if_no_path setting.
  */
@@ -1269,6 +1289,10 @@ static void multipath_resume(struct dm_target *ti)
 	struct multipath *m = (struct multipath *) ti->private;
 	unsigned long flags;
 
+	mutex_lock(&m->work_mutex);
+	m->suspended = 0;
+	mutex_unlock(&m->work_mutex);
+
 	spin_lock_irqsave(&m->lock, flags);
 	m->queue_if_no_path = m->saved_queue_if_no_path;
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -1397,51 +1421,66 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 {
-	int r;
+	int r = -EINVAL;
 	struct dm_dev *dev;
 	struct multipath *m = (struct multipath *) ti->private;
 	action_fn action;
 
+	mutex_lock(&m->work_mutex);
+
+	if (m->suspended) {
+		r = -EBUSY;
+		goto out;
+	}
+
 	if (argc == 1) {
-		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path")))
-			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, 0);
+		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path"))) {
+			r = queue_if_no_path(m, 1, 0);
+			goto out;
+		} else if (!strnicmp(argv[0], MESG_STR("fail_if_no_path"))) {
+			r = queue_if_no_path(m, 0, 0);
+			goto out;
+		}
 	}
 
-	if (argc != 2)
-		goto error;
+	if (argc != 2) {
+		DMWARN("Unrecognised multipath message received.");
+		goto out;
+	}
 
-	if (!strnicmp(argv[0], MESG_STR("disable_group")))
-		return bypass_pg_num(m, argv[1], 1);
-	else if (!strnicmp(argv[0], MESG_STR("enable_group")))
-		return bypass_pg_num(m, argv[1], 0);
-	else if (!strnicmp(argv[0], MESG_STR("switch_group")))
-		return switch_pg_num(m, argv[1]);
-	else if (!strnicmp(argv[0], MESG_STR("reinstate_path")))
+	if (!strnicmp(argv[0], MESG_STR("disable_group"))) {
+		r = bypass_pg_num(m, argv[1], 1);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("enable_group"))) {
+		r = bypass_pg_num(m, argv[1], 0);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("switch_group"))) {
+		r = switch_pg_num(m, argv[1]);
+		goto out;
+	} else if (!strnicmp(argv[0], MESG_STR("reinstate_path")))
 		action = reinstate_path;
 	else if (!strnicmp(argv[0], MESG_STR("fail_path")))
 		action = fail_path;
-	else
-		goto error;
+	else {
+		DMWARN("Unrecognised multipath message received.");
+		goto out;
+	}
 
 	r = dm_get_device(ti, argv[1], ti->begin, ti->len,
 			  dm_table_get_mode(ti->table), &dev);
 	if (r) {
 		DMWARN("message: error getting device %s",
 		       argv[1]);
-		return -EINVAL;
+		goto out;
 	}
 
 	r = action_dev(m, dev, action);
 
 	dm_put_device(ti, dev);
 
+out:
+	mutex_unlock(&m->work_mutex);
 	return r;
-
-error:
-	DMWARN("Unrecognised multipath message received.");
-	return -EINVAL;
 }
 
 static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
@@ -1567,13 +1606,14 @@ out:
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 1, 0},
+	.version = {1, 1, 1},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,
 	.map_rq = multipath_map,
 	.rq_end_io = multipath_end_io,
 	.presuspend = multipath_presuspend,
+	.postsuspend = multipath_postsuspend,
 	.resume = multipath_resume,
 	.status = multipath_status,
 	.message = multipath_message,
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 0c746420c008..157999ebd236 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -214,7 +214,7 @@ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, int rw,
 		    int metadata)
 {
 	struct dm_io_region where = {
-		.bdev = ps->store->cow->bdev,
+		.bdev = dm_snap_cow(ps->store->snap)->bdev,
 		.sector = ps->store->chunk_size * chunk,
 		.count = ps->store->chunk_size,
 	};
@@ -294,7 +294,8 @@ static int read_header(struct pstore *ps, int *new_snapshot)
 	 */
 	if (!ps->store->chunk_size) {
 		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
-		    bdev_logical_block_size(ps->store->cow->bdev) >> 9);
+		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
+					    bdev) >> 9);
 		ps->store->chunk_mask = ps->store->chunk_size - 1;
 		ps->store->chunk_shift = ffs(ps->store->chunk_size) - 1;
 		chunk_size_supplied = 0;
@@ -489,11 +490,22 @@ static struct pstore *get_info(struct dm_exception_store *store)
 	return (struct pstore *) store->context;
 }
 
-static void persistent_fraction_full(struct dm_exception_store *store,
-				     sector_t *numerator, sector_t *denominator)
+static void persistent_usage(struct dm_exception_store *store,
+			     sector_t *total_sectors,
+			     sector_t *sectors_allocated,
+			     sector_t *metadata_sectors)
 {
-	*numerator = get_info(store)->next_free * store->chunk_size;
-	*denominator = get_dev_size(store->cow->bdev);
+	struct pstore *ps = get_info(store);
+
+	*sectors_allocated = ps->next_free * store->chunk_size;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+
+	/*
+	 * First chunk is the fixed header.
+	 * Then there are (ps->current_area + 1) metadata chunks, each one
+	 * separated from the next by ps->exceptions_per_area data chunks.
+	 */
+	*metadata_sectors = (ps->current_area + 2) * store->chunk_size;
 }
 
 static void persistent_dtr(struct dm_exception_store *store)
@@ -552,44 +564,40 @@ static int persistent_read_metadata(struct dm_exception_store *store,
 		ps->current_area = 0;
 		zero_memory_area(ps);
 		r = zero_disk_area(ps, 0);
-		if (r) {
+		if (r)
 			DMWARN("zero_disk_area(0) failed");
-			return r;
-		}
-	} else {
-		/*
-		 * Sanity checks.
-		 */
-		if (ps->version != SNAPSHOT_DISK_VERSION) {
-			DMWARN("unable to handle snapshot disk version %d",
-			       ps->version);
-			return -EINVAL;
-		}
+		return r;
+	}
+	/*
+	 * Sanity checks.
+	 */
+	if (ps->version != SNAPSHOT_DISK_VERSION) {
+		DMWARN("unable to handle snapshot disk version %d",
+		       ps->version);
+		return -EINVAL;
+	}
 
-		/*
-		 * Metadata are valid, but snapshot is invalidated
-		 */
-		if (!ps->valid)
-			return 1;
+	/*
+	 * Metadata are valid, but snapshot is invalidated
+	 */
+	if (!ps->valid)
+		return 1;
 
-		/*
-		 * Read the metadata.
-		 */
-		r = read_exceptions(ps, callback, callback_context);
-		if (r)
-			return r;
-	}
+	/*
+	 * Read the metadata.
+	 */
+	r = read_exceptions(ps, callback, callback_context);
 
-	return 0;
+	return r;
 }
 
 static int persistent_prepare_exception(struct dm_exception_store *store,
-					struct dm_snap_exception *e)
+					struct dm_exception *e)
 {
 	struct pstore *ps = get_info(store);
 	uint32_t stride;
 	chunk_t next_free;
-	sector_t size = get_dev_size(store->cow->bdev);
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
 
 	/* Is there enough room ? */
 	if (size < ((ps->next_free + 1) * store->chunk_size))
@@ -611,7 +619,7 @@ static int persistent_prepare_exception(struct dm_exception_store *store,
 }
 
 static void persistent_commit_exception(struct dm_exception_store *store,
-					struct dm_snap_exception *e,
+					struct dm_exception *e,
 					void (*callback) (void *, int success),
 					void *callback_context)
 {
@@ -726,8 +734,7 @@ static unsigned persistent_status(struct dm_exception_store *store,
 	case STATUSTYPE_INFO:
 		break;
 	case STATUSTYPE_TABLE:
-		DMEMIT(" %s P %llu", store->cow->name,
-		       (unsigned long long)store->chunk_size);
+		DMEMIT(" P %llu", (unsigned long long)store->chunk_size);
 	}
 
 	return sz;
@@ -742,7 +749,7 @@ static struct dm_exception_store_type _persistent_type = {
 	.prepare_exception = persistent_prepare_exception,
 	.commit_exception = persistent_commit_exception,
 	.drop_snapshot = persistent_drop_snapshot,
-	.fraction_full = persistent_fraction_full,
+	.usage = persistent_usage,
 	.status = persistent_status,
 };
 
@@ -755,7 +762,7 @@ static struct dm_exception_store_type _persistent_compat_type = {
 	.prepare_exception = persistent_prepare_exception,
 	.commit_exception = persistent_commit_exception,
 	.drop_snapshot = persistent_drop_snapshot,
-	.fraction_full = persistent_fraction_full,
+	.usage = persistent_usage,
 	.status = persistent_status,
 };
 
diff --git a/drivers/md/dm-snap-transient.c b/drivers/md/dm-snap-transient.c
index cde5aa558e6d..a0898a66a2f8 100644
--- a/drivers/md/dm-snap-transient.c
+++ b/drivers/md/dm-snap-transient.c
@@ -36,10 +36,10 @@ static int transient_read_metadata(struct dm_exception_store *store,
 }
 
 static int transient_prepare_exception(struct dm_exception_store *store,
-				       struct dm_snap_exception *e)
+				       struct dm_exception *e)
 {
 	struct transient_c *tc = store->context;
-	sector_t size = get_dev_size(store->cow->bdev);
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
 
 	if (size < (tc->next_free + store->chunk_size))
 		return -1;
@@ -51,7 +51,7 @@ static int transient_prepare_exception(struct dm_exception_store *store,
 }
 
 static void transient_commit_exception(struct dm_exception_store *store,
-				       struct dm_snap_exception *e,
+				       struct dm_exception *e,
 				       void (*callback) (void *, int success),
 				       void *callback_context)
 {
@@ -59,11 +59,14 @@ static void transient_commit_exception(struct dm_exception_store *store,
 	callback(callback_context, 1);
 }
 
-static void transient_fraction_full(struct dm_exception_store *store,
-				    sector_t *numerator, sector_t *denominator)
+static void transient_usage(struct dm_exception_store *store,
+			    sector_t *total_sectors,
+			    sector_t *sectors_allocated,
+			    sector_t *metadata_sectors)
 {
-	*numerator = ((struct transient_c *) store->context)->next_free;
-	*denominator = get_dev_size(store->cow->bdev);
+	*sectors_allocated = ((struct transient_c *) store->context)->next_free;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+	*metadata_sectors = 0;
 }
 
 static int transient_ctr(struct dm_exception_store *store,
@@ -91,8 +94,7 @@ static unsigned transient_status(struct dm_exception_store *store,
 	case STATUSTYPE_INFO:
 		break;
 	case STATUSTYPE_TABLE:
-		DMEMIT(" %s N %llu", store->cow->name,
-		       (unsigned long long)store->chunk_size);
+		DMEMIT(" N %llu", (unsigned long long)store->chunk_size);
 	}
 
 	return sz;
@@ -106,7 +108,7 @@ static struct dm_exception_store_type _transient_type = {
 	.read_metadata = transient_read_metadata,
 	.prepare_exception = transient_prepare_exception,
 	.commit_exception = transient_commit_exception,
-	.fraction_full = transient_fraction_full,
+	.usage = transient_usage,
 	.status = transient_status,
 };
 
@@ -118,7 +120,7 @@ static struct dm_exception_store_type _transient_compat_type = {
 	.read_metadata = transient_read_metadata,
 	.prepare_exception = transient_prepare_exception,
 	.commit_exception = transient_commit_exception,
-	.fraction_full = transient_fraction_full,
+	.usage = transient_usage,
 	.status = transient_status,
 };
 
diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c
index 3a3ba46e6d4b..c6c26a38c845 100644
--- a/drivers/md/dm-snap.c
+++ b/drivers/md/dm-snap.c
@@ -49,7 +49,7 @@
 #define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
 					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
 
-struct exception_table {
+struct dm_exception_table {
 	uint32_t hash_mask;
 	unsigned hash_shift;
 	struct list_head *table;
@@ -59,6 +59,9 @@ struct dm_snapshot {
 	struct rw_semaphore lock;
 
 	struct dm_dev *origin;
+	struct dm_dev *cow;
+
+	struct dm_target *ti;
 
 	/* List of snapshots per Origin */
 	struct list_head list;
@@ -69,12 +72,15 @@ struct dm_snapshot {
 	/* Origin writes don't trigger exceptions until this is set */
 	int active;
 
+	/* Whether or not owning mapped_device is suspended */
+	int suspended;
+
 	mempool_t *pending_pool;
 
 	atomic_t pending_exceptions_count;
 
-	struct exception_table pending;
-	struct exception_table complete;
+	struct dm_exception_table pending;
+	struct dm_exception_table complete;
 
 	/*
 	 * pe_lock protects all pending_exception operations and access
@@ -97,6 +103,12 @@ struct dm_snapshot {
 	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
 };
 
+struct dm_dev *dm_snap_cow(struct dm_snapshot *s)
+{
+	return s->cow;
+}
+EXPORT_SYMBOL(dm_snap_cow);
+
 static struct workqueue_struct *ksnapd;
 static void flush_queued_bios(struct work_struct *work);
 
@@ -116,7 +128,7 @@ static int bdev_equal(struct block_device *lhs, struct block_device *rhs)
 }
 
 struct dm_snap_pending_exception {
-	struct dm_snap_exception e;
+	struct dm_exception e;
 
 	/*
 	 * Origin buffers waiting for this to complete are held
@@ -351,8 +363,8 @@ static void unregister_snapshot(struct dm_snapshot *s)
  * The lowest hash_shift bits of the chunk number are ignored, allowing
  * some consecutive chunks to be grouped together.
  */
-static int init_exception_table(struct exception_table *et, uint32_t size,
-				unsigned hash_shift)
+static int dm_exception_table_init(struct dm_exception_table *et,
+				   uint32_t size, unsigned hash_shift)
 {
 	unsigned int i;
 
@@ -368,10 +380,11 @@ static int init_exception_table(struct exception_table *et, uint32_t size,
 	return 0;
 }
 
-static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem)
+static void dm_exception_table_exit(struct dm_exception_table *et,
+				    struct kmem_cache *mem)
 {
 	struct list_head *slot;
-	struct dm_snap_exception *ex, *next;
+	struct dm_exception *ex, *next;
 	int i, size;
 
 	size = et->hash_mask + 1;
@@ -385,19 +398,12 @@ static void exit_exception_table(struct exception_table *et, struct kmem_cache *
 	vfree(et->table);
 }
 
-static uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
+static uint32_t exception_hash(struct dm_exception_table *et, chunk_t chunk)
 {
 	return (chunk >> et->hash_shift) & et->hash_mask;
 }
 
-static void insert_exception(struct exception_table *eh,
-			     struct dm_snap_exception *e)
-{
-	struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)];
-	list_add(&e->hash_list, l);
-}
-
-static void remove_exception(struct dm_snap_exception *e)
+static void dm_remove_exception(struct dm_exception *e)
 {
 	list_del(&e->hash_list);
 }
@@ -406,11 +412,11 @@ static void remove_exception(struct dm_snap_exception *e)
  * Return the exception data for a sector, or NULL if not
  * remapped.
  */
-static struct dm_snap_exception *lookup_exception(struct exception_table *et,
-						  chunk_t chunk)
+static struct dm_exception *dm_lookup_exception(struct dm_exception_table *et,
+						chunk_t chunk)
 {
 	struct list_head *slot;
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 
 	slot = &et->table[exception_hash(et, chunk)];
 	list_for_each_entry (e, slot, hash_list)
@@ -421,9 +427,9 @@ static struct dm_snap_exception *lookup_exception(struct exception_table *et,
 	return NULL;
 }
 
-static struct dm_snap_exception *alloc_exception(void)
+static struct dm_exception *alloc_completed_exception(void)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 
 	e = kmem_cache_alloc(exception_cache, GFP_NOIO);
 	if (!e)
@@ -432,7 +438,7 @@ static struct dm_snap_exception *alloc_exception(void)
 	return e;
 }
 
-static void free_exception(struct dm_snap_exception *e)
+static void free_completed_exception(struct dm_exception *e)
 {
 	kmem_cache_free(exception_cache, e);
 }
@@ -457,12 +463,11 @@ static void free_pending_exception(struct dm_snap_pending_exception *pe)
 	atomic_dec(&s->pending_exceptions_count);
 }
 
-static void insert_completed_exception(struct dm_snapshot *s,
-				       struct dm_snap_exception *new_e)
+static void dm_insert_exception(struct dm_exception_table *eh,
+				struct dm_exception *new_e)
 {
-	struct exception_table *eh = &s->complete;
 	struct list_head *l;
-	struct dm_snap_exception *e = NULL;
+	struct dm_exception *e = NULL;
 
 	l = &eh->table[exception_hash(eh, new_e->old_chunk)];
 
@@ -478,7 +483,7 @@ static void insert_completed_exception(struct dm_snapshot *s,
 		    new_e->new_chunk == (dm_chunk_number(e->new_chunk) +
 					 dm_consecutive_chunk_count(e) + 1)) {
 			dm_consecutive_chunk_count_inc(e);
-			free_exception(new_e);
+			free_completed_exception(new_e);
 			return;
 		}
 
@@ -488,7 +493,7 @@ static void insert_completed_exception(struct dm_snapshot *s,
 			dm_consecutive_chunk_count_inc(e);
 			e->old_chunk--;
 			e->new_chunk--;
-			free_exception(new_e);
+			free_completed_exception(new_e);
 			return;
 		}
 
@@ -507,9 +512,9 @@ out:
 static int dm_add_exception(void *context, chunk_t old, chunk_t new)
 {
 	struct dm_snapshot *s = context;
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 
-	e = alloc_exception();
+	e = alloc_completed_exception();
 	if (!e)
 		return -ENOMEM;
 
@@ -518,11 +523,30 @@ static int dm_add_exception(void *context, chunk_t old, chunk_t new)
 	/* Consecutive_count is implicitly initialised to zero */
 	e->new_chunk = new;
 
-	insert_completed_exception(s, e);
+	dm_insert_exception(&s->complete, e);
 
 	return 0;
 }
 
+#define min_not_zero(l, r) (((l) == 0) ? (r) : (((r) == 0) ? (l) : min(l, r)))
+
+/*
+ * Return a minimum chunk size of all snapshots that have the specified origin.
+ * Return zero if the origin has no snapshots.
+ */
+static sector_t __minimum_chunk_size(struct origin *o)
+{
+	struct dm_snapshot *snap;
+	unsigned chunk_size = 0;
+
+	if (o)
+		list_for_each_entry(snap, &o->snapshots, list)
+			chunk_size = min_not_zero(chunk_size,
+						  snap->store->chunk_size);
+
+	return chunk_size;
+}
+
 /*
  * Hard coded magic.
  */
@@ -546,7 +570,7 @@ static int init_hash_tables(struct dm_snapshot *s)
 	 * Calculate based on the size of the original volume or
 	 * the COW volume...
 	 */
-	cow_dev_size = get_dev_size(s->store->cow->bdev);
+	cow_dev_size = get_dev_size(s->cow->bdev);
 	origin_dev_size = get_dev_size(s->origin->bdev);
 	max_buckets = calc_max_buckets();
 
@@ -554,8 +578,8 @@ static int init_hash_tables(struct dm_snapshot *s)
 	hash_size = min(hash_size, max_buckets);
 
 	hash_size = rounddown_pow_of_two(hash_size);
-	if (init_exception_table(&s->complete, hash_size,
-				 DM_CHUNK_CONSECUTIVE_BITS))
+	if (dm_exception_table_init(&s->complete, hash_size,
+				    DM_CHUNK_CONSECUTIVE_BITS))
 		return -ENOMEM;
 
 	/*
@@ -566,8 +590,8 @@ static int init_hash_tables(struct dm_snapshot *s)
 	if (hash_size < 64)
 		hash_size = 64;
 
-	if (init_exception_table(&s->pending, hash_size, 0)) {
-		exit_exception_table(&s->complete, exception_cache);
+	if (dm_exception_table_init(&s->pending, hash_size, 0)) {
+		dm_exception_table_exit(&s->complete, exception_cache);
 		return -ENOMEM;
 	}
 
@@ -582,47 +606,58 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	struct dm_snapshot *s;
 	int i;
 	int r = -EINVAL;
-	char *origin_path;
-	struct dm_exception_store *store;
+	char *origin_path, *cow_path;
 	unsigned args_used;
 
 	if (argc != 4) {
 		ti->error = "requires exactly 4 arguments";
 		r = -EINVAL;
-		goto bad_args;
+		goto bad;
 	}
 
 	origin_path = argv[0];
 	argv++;
 	argc--;
 
-	r = dm_exception_store_create(ti, argc, argv, &args_used, &store);
+	s = kmalloc(sizeof(*s), GFP_KERNEL);
+	if (!s) {
+		ti->error = "Cannot allocate snapshot context private "
+		    "structure";
+		r = -ENOMEM;
+		goto bad;
+	}
+
+	cow_path = argv[0];
+	argv++;
+	argc--;
+
+	r = dm_get_device(ti, cow_path, 0, 0,
+			  FMODE_READ | FMODE_WRITE, &s->cow);
+	if (r) {
+		ti->error = "Cannot get COW device";
+		goto bad_cow;
+	}
+
+	r = dm_exception_store_create(ti, argc, argv, s, &args_used, &s->store);
 	if (r) {
 		ti->error = "Couldn't create exception store";
 		r = -EINVAL;
-		goto bad_args;
+		goto bad_store;
 	}
 
 	argv += args_used;
 	argc -= args_used;
 
-	s = kmalloc(sizeof(*s), GFP_KERNEL);
-	if (!s) {
-		ti->error = "Cannot allocate snapshot context private "
-		    "structure";
-		r = -ENOMEM;
-		goto bad_snap;
-	}
-
 	r = dm_get_device(ti, origin_path, 0, ti->len, FMODE_READ, &s->origin);
 	if (r) {
 		ti->error = "Cannot get origin device";
 		goto bad_origin;
 	}
 
-	s->store = store;
+	s->ti = ti;
 	s->valid = 1;
 	s->active = 0;
+	s->suspended = 0;
 	atomic_set(&s->pending_exceptions_count, 0);
 	init_rwsem(&s->lock);
 	spin_lock_init(&s->pe_lock);
@@ -702,19 +737,22 @@ bad_pending_pool:
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 
 bad_kcopyd:
-	exit_exception_table(&s->pending, pending_cache);
-	exit_exception_table(&s->complete, exception_cache);
+	dm_exception_table_exit(&s->pending, pending_cache);
+	dm_exception_table_exit(&s->complete, exception_cache);
 
 bad_hash_tables:
 	dm_put_device(ti, s->origin);
 
 bad_origin:
-	kfree(s);
+	dm_exception_store_destroy(s->store);
+
+bad_store:
+	dm_put_device(ti, s->cow);
 
-bad_snap:
-	dm_exception_store_destroy(store);
+bad_cow:
+	kfree(s);
 
-bad_args:
+bad:
 	return r;
 }
 
@@ -723,8 +761,8 @@ static void __free_exceptions(struct dm_snapshot *s)
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 	s->kcopyd_client = NULL;
 
-	exit_exception_table(&s->pending, pending_cache);
-	exit_exception_table(&s->complete, exception_cache);
+	dm_exception_table_exit(&s->pending, pending_cache);
+	dm_exception_table_exit(&s->complete, exception_cache);
 }
 
 static void snapshot_dtr(struct dm_target *ti)
@@ -763,6 +801,8 @@ static void snapshot_dtr(struct dm_target *ti)
 
 	dm_exception_store_destroy(s->store);
 
+	dm_put_device(ti, s->cow);
+
 	kfree(s);
 }
 
@@ -825,7 +865,7 @@ static void __invalidate_snapshot(struct dm_snapshot *s, int err)
 
 	s->valid = 0;
 
-	dm_table_event(s->store->ti->table);
+	dm_table_event(s->ti->table);
 }
 
 static void get_pending_exception(struct dm_snap_pending_exception *pe)
@@ -863,7 +903,7 @@ static struct bio *put_pending_exception(struct dm_snap_pending_exception *pe)
 
 static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 	struct dm_snapshot *s = pe->snap;
 	struct bio *origin_bios = NULL;
 	struct bio *snapshot_bios = NULL;
@@ -877,7 +917,7 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 		goto out;
 	}
 
-	e = alloc_exception();
+	e = alloc_completed_exception();
 	if (!e) {
 		down_write(&s->lock);
 		__invalidate_snapshot(s, -ENOMEM);
@@ -888,7 +928,7 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 
 	down_write(&s->lock);
 	if (!s->valid) {
-		free_exception(e);
+		free_completed_exception(e);
 		error = 1;
 		goto out;
 	}
@@ -904,10 +944,10 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 	 * Add a proper exception, and remove the
 	 * in-flight exception from the list.
 	 */
-	insert_completed_exception(s, e);
+	dm_insert_exception(&s->complete, e);
 
  out:
-	remove_exception(&pe->e);
+	dm_remove_exception(&pe->e);
 	snapshot_bios = bio_list_get(&pe->snapshot_bios);
 	origin_bios = put_pending_exception(pe);
 
@@ -963,7 +1003,7 @@ static void start_copy(struct dm_snap_pending_exception *pe)
 	src.sector = chunk_to_sector(s->store, pe->e.old_chunk);
 	src.count = min((sector_t)s->store->chunk_size, dev_size - src.sector);
 
-	dest.bdev = s->store->cow->bdev;
+	dest.bdev = s->cow->bdev;
 	dest.sector = chunk_to_sector(s->store, pe->e.new_chunk);
 	dest.count = src.count;
 
@@ -975,7 +1015,7 @@ static void start_copy(struct dm_snap_pending_exception *pe)
 static struct dm_snap_pending_exception *
 __lookup_pending_exception(struct dm_snapshot *s, chunk_t chunk)
 {
-	struct dm_snap_exception *e = lookup_exception(&s->pending, chunk);
+	struct dm_exception *e = dm_lookup_exception(&s->pending, chunk);
 
 	if (!e)
 		return NULL;
@@ -1016,15 +1056,15 @@ __find_pending_exception(struct dm_snapshot *s,
 	}
 
 	get_pending_exception(pe);
-	insert_exception(&s->pending, &pe->e);
+	dm_insert_exception(&s->pending, &pe->e);
 
 	return pe;
 }
 
-static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
+static void remap_exception(struct dm_snapshot *s, struct dm_exception *e,
 			    struct bio *bio, chunk_t chunk)
 {
-	bio->bi_bdev = s->store->cow->bdev;
+	bio->bi_bdev = s->cow->bdev;
 	bio->bi_sector = chunk_to_sector(s->store,
 					 dm_chunk_number(e->new_chunk) +
 					 (chunk - e->old_chunk)) +
@@ -1035,14 +1075,14 @@ static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
 static int snapshot_map(struct dm_target *ti, struct bio *bio,
 			union map_info *map_context)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 	struct dm_snapshot *s = ti->private;
 	int r = DM_MAPIO_REMAPPED;
 	chunk_t chunk;
 	struct dm_snap_pending_exception *pe = NULL;
 
 	if (unlikely(bio_empty_barrier(bio))) {
-		bio->bi_bdev = s->store->cow->bdev;
+		bio->bi_bdev = s->cow->bdev;
 		return DM_MAPIO_REMAPPED;
 	}
 
@@ -1063,7 +1103,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	}
 
 	/* If the block is already remapped - use that, else remap it */
-	e = lookup_exception(&s->complete, chunk);
+	e = dm_lookup_exception(&s->complete, chunk);
 	if (e) {
 		remap_exception(s, e, bio, chunk);
 		goto out_unlock;
@@ -1087,7 +1127,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 				goto out_unlock;
 			}
 
-			e = lookup_exception(&s->complete, chunk);
+			e = dm_lookup_exception(&s->complete, chunk);
 			if (e) {
 				free_pending_exception(pe);
 				remap_exception(s, e, bio, chunk);
@@ -1137,12 +1177,22 @@ static int snapshot_end_io(struct dm_target *ti, struct bio *bio,
 	return 0;
 }
 
+static void snapshot_postsuspend(struct dm_target *ti)
+{
+	struct dm_snapshot *s = ti->private;
+
+	down_write(&s->lock);
+	s->suspended = 1;
+	up_write(&s->lock);
+}
+
 static void snapshot_resume(struct dm_target *ti)
 {
 	struct dm_snapshot *s = ti->private;
 
 	down_write(&s->lock);
 	s->active = 1;
+	s->suspended = 0;
 	up_write(&s->lock);
 }
 
@@ -1159,14 +1209,17 @@ static int snapshot_status(struct dm_target *ti, status_type_t type,
 		if (!snap->valid)
 			DMEMIT("Invalid");
 		else {
-			if (snap->store->type->fraction_full) {
-				sector_t numerator, denominator;
-				snap->store->type->fraction_full(snap->store,
-								 &numerator,
-								 &denominator);
-				DMEMIT("%llu/%llu",
-				       (unsigned long long)numerator,
-				       (unsigned long long)denominator);
+			if (snap->store->type->usage) {
+				sector_t total_sectors, sectors_allocated,
+					 metadata_sectors;
+				snap->store->type->usage(snap->store,
+							 &total_sectors,
+							 &sectors_allocated,
+							 &metadata_sectors);
+				DMEMIT("%llu/%llu %llu",
+				       (unsigned long long)sectors_allocated,
+				       (unsigned long long)total_sectors,
+				       (unsigned long long)metadata_sectors);
 			}
 			else
 				DMEMIT("Unknown");
@@ -1179,7 +1232,7 @@ static int snapshot_status(struct dm_target *ti, status_type_t type,
 		 * to make private copies if the output is to
 		 * make sense.
 		 */
-		DMEMIT("%s", snap->origin->name);
+		DMEMIT("%s %s", snap->origin->name, snap->cow->name);
 		snap->store->type->status(snap->store, type, result + sz,
 					  maxlen - sz);
 		break;
@@ -1206,7 +1259,7 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 {
 	int r = DM_MAPIO_REMAPPED, first = 0;
 	struct dm_snapshot *snap;
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 	struct dm_snap_pending_exception *pe, *next_pe, *primary_pe = NULL;
 	chunk_t chunk;
 	LIST_HEAD(pe_queue);
@@ -1221,7 +1274,7 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 			goto next_snapshot;
 
 		/* Nothing to do if writing beyond end of snapshot */
-		if (bio->bi_sector >= dm_table_get_size(snap->store->ti->table))
+		if (bio->bi_sector >= dm_table_get_size(snap->ti->table))
 			goto next_snapshot;
 
 		/*
@@ -1238,7 +1291,7 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 		 * ref_count is initialised to 1 so pending_complete()
 		 * won't destroy the primary_pe while we're inside this loop.
 		 */
-		e = lookup_exception(&snap->complete, chunk);
+		e = dm_lookup_exception(&snap->complete, chunk);
 		if (e)
 			goto next_snapshot;
 
@@ -1253,7 +1306,7 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 				goto next_snapshot;
 			}
 
-			e = lookup_exception(&snap->complete, chunk);
+			e = dm_lookup_exception(&snap->complete, chunk);
 			if (e) {
 				free_pending_exception(pe);
 				goto next_snapshot;
@@ -1391,8 +1444,6 @@ static int origin_map(struct dm_target *ti, struct bio *bio,
 	return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED;
 }
 
-#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))
-
 /*
  * Set the target "split_io" field to the minimum of all the snapshots'
  * chunk sizes.
@@ -1400,19 +1451,12 @@ static int origin_map(struct dm_target *ti, struct bio *bio,
 static void origin_resume(struct dm_target *ti)
 {
 	struct dm_dev *dev = ti->private;
-	struct dm_snapshot *snap;
-	struct origin *o;
-	unsigned chunk_size = 0;
 
 	down_read(&_origins_lock);
-	o = __lookup_origin(dev->bdev);
-	if (o)
-		list_for_each_entry (snap, &o->snapshots, list)
-			chunk_size = min_not_zero(chunk_size,
-						  snap->store->chunk_size);
-	up_read(&_origins_lock);
 
-	ti->split_io = chunk_size;
+	ti->split_io = __minimum_chunk_size(__lookup_origin(dev->bdev));
+
+	up_read(&_origins_lock);
 }
 
 static int origin_status(struct dm_target *ti, status_type_t type, char *result,
@@ -1455,12 +1499,13 @@ static struct target_type origin_target = {
 
 static struct target_type snapshot_target = {
 	.name    = "snapshot",
-	.version = {1, 7, 0},
+	.version = {1, 9, 0},
 	.module  = THIS_MODULE,
 	.ctr     = snapshot_ctr,
 	.dtr     = snapshot_dtr,
 	.map     = snapshot_map,
 	.end_io  = snapshot_end_io,
+	.postsuspend = snapshot_postsuspend,
 	.resume  = snapshot_resume,
 	.status  = snapshot_status,
 	.iterate_devices = snapshot_iterate_devices,
@@ -1494,7 +1539,7 @@ static int __init dm_snapshot_init(void)
 		goto bad2;
 	}
 
-	exception_cache = KMEM_CACHE(dm_snap_exception, 0);
+	exception_cache = KMEM_CACHE(dm_exception, 0);
 	if (!exception_cache) {
 		DMERR("Couldn't create exception cache.");
 		r = -ENOMEM;
diff --git a/drivers/md/dm-sysfs.c b/drivers/md/dm-sysfs.c
index 4b045903a4e2..b000de38c99a 100644
--- a/drivers/md/dm-sysfs.c
+++ b/drivers/md/dm-sysfs.c
@@ -80,12 +80,20 @@ static struct sysfs_ops dm_sysfs_ops = {
 };
 
 /*
+ * The sysfs structure is embedded in md struct, nothing to do here
+ */
+static void dm_sysfs_release(struct kobject *kobj)
+{
+}
+
+/*
  * dm kobject is embedded in mapped_device structure
  * no need to define release function here
  */
 static struct kobj_type dm_ktype = {
 	.sysfs_ops	= &dm_sysfs_ops,
 	.default_attrs	= dm_attrs,
+	.release	= dm_sysfs_release
 };
 
 /*
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 1a6cb3c7822e..3162b9c3c3f2 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -237,6 +237,9 @@ void dm_table_destroy(struct dm_table *t)
 {
 	unsigned int i;
 
+	if (!t)
+		return;
+
 	while (atomic_read(&t->holders))
 		msleep(1);
 	smp_mb();
diff --git a/drivers/md/dm-uevent.c b/drivers/md/dm-uevent.c
index 6f65883aef12..c7c555a8c7b2 100644
--- a/drivers/md/dm-uevent.c
+++ b/drivers/md/dm-uevent.c
@@ -139,14 +139,13 @@ void dm_send_uevents(struct list_head *events, struct kobject *kobj)
 		list_del_init(&event->elist);
 
 		/*
-		 * Need to call dm_copy_name_and_uuid from here for now.
-		 * Context of previous var adds and locking used for
-		 * hash_cell not compatable.
+		 * When a device is being removed this copy fails and we
+		 * discard these unsent events.
 		 */
 		if (dm_copy_name_and_uuid(event->md, event->name,
 					  event->uuid)) {
-			DMERR("%s: dm_copy_name_and_uuid() failed",
-			      __func__);
+			DMINFO("%s: skipping sending uevent for lost device",
+			       __func__);
 			goto uevent_free;
 		}
 
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 724efc63904d..c254c6cf69a1 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -143,9 +143,19 @@ struct mapped_device {
 	int barrier_error;
 
 	/*
+	 * Protect barrier_error from concurrent endio processing
+	 * in request-based dm.
+	 */
+	spinlock_t barrier_error_lock;
+
+	/*
 	 * Processing queue (flush/barriers)
 	 */
 	struct workqueue_struct *wq;
+	struct work_struct barrier_work;
+
+	/* A pointer to the currently processing pre/post flush request */
+	struct request *flush_request;
 
 	/*
 	 * The current mapping.
@@ -178,9 +188,6 @@ struct mapped_device {
 	/* forced geometry settings */
 	struct hd_geometry geometry;
 
-	/* marker of flush suspend for request-based dm */
-	struct request suspend_rq;
-
 	/* For saving the address of __make_request for request based dm */
 	make_request_fn *saved_make_request_fn;
 
@@ -275,6 +282,7 @@ static int (*_inits[])(void) __initdata = {
 	dm_target_init,
 	dm_linear_init,
 	dm_stripe_init,
+	dm_io_init,
 	dm_kcopyd_init,
 	dm_interface_init,
 };
@@ -284,6 +292,7 @@ static void (*_exits[])(void) = {
 	dm_target_exit,
 	dm_linear_exit,
 	dm_stripe_exit,
+	dm_io_exit,
 	dm_kcopyd_exit,
 	dm_interface_exit,
 };
@@ -320,6 +329,11 @@ static void __exit dm_exit(void)
 /*
  * Block device functions
  */
+int dm_deleting_md(struct mapped_device *md)
+{
+	return test_bit(DMF_DELETING, &md->flags);
+}
+
 static int dm_blk_open(struct block_device *bdev, fmode_t mode)
 {
 	struct mapped_device *md;
@@ -331,7 +345,7 @@ static int dm_blk_open(struct block_device *bdev, fmode_t mode)
 		goto out;
 
 	if (test_bit(DMF_FREEING, &md->flags) ||
-	    test_bit(DMF_DELETING, &md->flags)) {
+	    dm_deleting_md(md)) {
 		md = NULL;
 		goto out;
 	}
@@ -388,7 +402,7 @@ static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
 			unsigned int cmd, unsigned long arg)
 {
 	struct mapped_device *md = bdev->bd_disk->private_data;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table(md);
 	struct dm_target *tgt;
 	int r = -ENOTTY;
 
@@ -430,9 +444,10 @@ static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
 	mempool_free(tio, md->tio_pool);
 }
 
-static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md)
+static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
+					    gfp_t gfp_mask)
 {
-	return mempool_alloc(md->tio_pool, GFP_ATOMIC);
+	return mempool_alloc(md->tio_pool, gfp_mask);
 }
 
 static void free_rq_tio(struct dm_rq_target_io *tio)
@@ -450,6 +465,12 @@ static void free_bio_info(struct dm_rq_clone_bio_info *info)
 	mempool_free(info, info->tio->md->io_pool);
 }
 
+static int md_in_flight(struct mapped_device *md)
+{
+	return atomic_read(&md->pending[READ]) +
+	       atomic_read(&md->pending[WRITE]);
+}
+
 static void start_io_acct(struct dm_io *io)
 {
 	struct mapped_device *md = io->md;
@@ -512,7 +533,7 @@ static void queue_io(struct mapped_device *md, struct bio *bio)
  * function to access the md->map field, and make sure they call
  * dm_table_put() when finished.
  */
-struct dm_table *dm_get_table(struct mapped_device *md)
+struct dm_table *dm_get_live_table(struct mapped_device *md)
 {
 	struct dm_table *t;
 	unsigned long flags;
@@ -716,28 +737,38 @@ static void end_clone_bio(struct bio *clone, int error)
 	blk_update_request(tio->orig, 0, nr_bytes);
 }
 
+static void store_barrier_error(struct mapped_device *md, int error)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&md->barrier_error_lock, flags);
+	/*
+	 * Basically, the first error is taken, but:
+	 *   -EOPNOTSUPP supersedes any I/O error.
+	 *   Requeue request supersedes any I/O error but -EOPNOTSUPP.
+	 */
+	if (!md->barrier_error || error == -EOPNOTSUPP ||
+	    (md->barrier_error != -EOPNOTSUPP &&
+	     error == DM_ENDIO_REQUEUE))
+		md->barrier_error = error;
+	spin_unlock_irqrestore(&md->barrier_error_lock, flags);
+}
+
 /*
  * Don't touch any member of the md after calling this function because
  * the md may be freed in dm_put() at the end of this function.
  * Or do dm_get() before calling this function and dm_put() later.
  */
-static void rq_completed(struct mapped_device *md, int run_queue)
+static void rq_completed(struct mapped_device *md, int rw, int run_queue)
 {
-	int wakeup_waiters = 0;
-	struct request_queue *q = md->queue;
-	unsigned long flags;
-
-	spin_lock_irqsave(q->queue_lock, flags);
-	if (!queue_in_flight(q))
-		wakeup_waiters = 1;
-	spin_unlock_irqrestore(q->queue_lock, flags);
+	atomic_dec(&md->pending[rw]);
 
 	/* nudge anyone waiting on suspend queue */
-	if (wakeup_waiters)
+	if (!md_in_flight(md))
 		wake_up(&md->wait);
 
 	if (run_queue)
-		blk_run_queue(q);
+		blk_run_queue(md->queue);
 
 	/*
 	 * dm_put() must be at the end of this function. See the comment above
@@ -753,6 +784,44 @@ static void free_rq_clone(struct request *clone)
 	free_rq_tio(tio);
 }
 
+/*
+ * Complete the clone and the original request.
+ * Must be called without queue lock.
+ */
+static void dm_end_request(struct request *clone, int error)
+{
+	int rw = rq_data_dir(clone);
+	int run_queue = 1;
+	bool is_barrier = blk_barrier_rq(clone);
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	struct mapped_device *md = tio->md;
+	struct request *rq = tio->orig;
+
+	if (blk_pc_request(rq) && !is_barrier) {
+		rq->errors = clone->errors;
+		rq->resid_len = clone->resid_len;
+
+		if (rq->sense)
+			/*
+			 * We are using the sense buffer of the original
+			 * request.
+			 * So setting the length of the sense data is enough.
+			 */
+			rq->sense_len = clone->sense_len;
+	}
+
+	free_rq_clone(clone);
+
+	if (unlikely(is_barrier)) {
+		if (unlikely(error))
+			store_barrier_error(md, error);
+		run_queue = 0;
+	} else
+		blk_end_request_all(rq, error);
+
+	rq_completed(md, rw, run_queue);
+}
+
 static void dm_unprep_request(struct request *rq)
 {
 	struct request *clone = rq->special;
@@ -768,12 +837,23 @@ static void dm_unprep_request(struct request *rq)
  */
 void dm_requeue_unmapped_request(struct request *clone)
 {
+	int rw = rq_data_dir(clone);
 	struct dm_rq_target_io *tio = clone->end_io_data;
 	struct mapped_device *md = tio->md;
 	struct request *rq = tio->orig;
 	struct request_queue *q = rq->q;
 	unsigned long flags;
 
+	if (unlikely(blk_barrier_rq(clone))) {
+		/*
+		 * Barrier clones share an original request.
+		 * Leave it to dm_end_request(), which handles this special
+		 * case.
+		 */
+		dm_end_request(clone, DM_ENDIO_REQUEUE);
+		return;
+	}
+
 	dm_unprep_request(rq);
 
 	spin_lock_irqsave(q->queue_lock, flags);
@@ -782,7 +862,7 @@ void dm_requeue_unmapped_request(struct request *clone)
 	blk_requeue_request(q, rq);
 	spin_unlock_irqrestore(q->queue_lock, flags);
 
-	rq_completed(md, 0);
+	rq_completed(md, rw, 0);
 }
 EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request);
 
@@ -815,34 +895,28 @@ static void start_queue(struct request_queue *q)
 	spin_unlock_irqrestore(q->queue_lock, flags);
 }
 
-/*
- * Complete the clone and the original request.
- * Must be called without queue lock.
- */
-static void dm_end_request(struct request *clone, int error)
+static void dm_done(struct request *clone, int error, bool mapped)
 {
+	int r = error;
 	struct dm_rq_target_io *tio = clone->end_io_data;
-	struct mapped_device *md = tio->md;
-	struct request *rq = tio->orig;
+	dm_request_endio_fn rq_end_io = tio->ti->type->rq_end_io;
 
-	if (blk_pc_request(rq)) {
-		rq->errors = clone->errors;
-		rq->resid_len = clone->resid_len;
+	if (mapped && rq_end_io)
+		r = rq_end_io(tio->ti, clone, error, &tio->info);
 
-		if (rq->sense)
-			/*
-			 * We are using the sense buffer of the original
-			 * request.
-			 * So setting the length of the sense data is enough.
-			 */
-			rq->sense_len = clone->sense_len;
+	if (r <= 0)
+		/* The target wants to complete the I/O */
+		dm_end_request(clone, r);
+	else if (r == DM_ENDIO_INCOMPLETE)
+		/* The target will handle the I/O */
+		return;
+	else if (r == DM_ENDIO_REQUEUE)
+		/* The target wants to requeue the I/O */
+		dm_requeue_unmapped_request(clone);
+	else {
+		DMWARN("unimplemented target endio return value: %d", r);
+		BUG();
 	}
-
-	free_rq_clone(clone);
-
-	blk_end_request_all(rq, error);
-
-	rq_completed(md, 1);
 }
 
 /*
@@ -850,27 +924,14 @@ static void dm_end_request(struct request *clone, int error)
  */
 static void dm_softirq_done(struct request *rq)
 {
+	bool mapped = true;
 	struct request *clone = rq->completion_data;
 	struct dm_rq_target_io *tio = clone->end_io_data;
-	dm_request_endio_fn rq_end_io = tio->ti->type->rq_end_io;
-	int error = tio->error;
 
-	if (!(rq->cmd_flags & REQ_FAILED) && rq_end_io)
-		error = rq_end_io(tio->ti, clone, error, &tio->info);
+	if (rq->cmd_flags & REQ_FAILED)
+		mapped = false;
 
-	if (error <= 0)
-		/* The target wants to complete the I/O */
-		dm_end_request(clone, error);
-	else if (error == DM_ENDIO_INCOMPLETE)
-		/* The target will handle the I/O */
-		return;
-	else if (error == DM_ENDIO_REQUEUE)
-		/* The target wants to requeue the I/O */
-		dm_requeue_unmapped_request(clone);
-	else {
-		DMWARN("unimplemented target endio return value: %d", error);
-		BUG();
-	}
+	dm_done(clone, tio->error, mapped);
 }
 
 /*
@@ -882,6 +943,19 @@ static void dm_complete_request(struct request *clone, int error)
 	struct dm_rq_target_io *tio = clone->end_io_data;
 	struct request *rq = tio->orig;
 
+	if (unlikely(blk_barrier_rq(clone))) {
+		/*
+		 * Barrier clones share an original request.  So can't use
+		 * softirq_done with the original.
+		 * Pass the clone to dm_done() directly in this special case.
+		 * It is safe (even if clone->q->queue_lock is held here)
+		 * because there is no I/O dispatching during the completion
+		 * of barrier clone.
+		 */
+		dm_done(clone, error, true);
+		return;
+	}
+
 	tio->error = error;
 	rq->completion_data = clone;
 	blk_complete_request(rq);
@@ -898,6 +972,17 @@ void dm_kill_unmapped_request(struct request *clone, int error)
 	struct dm_rq_target_io *tio = clone->end_io_data;
 	struct request *rq = tio->orig;
 
+	if (unlikely(blk_barrier_rq(clone))) {
+		/*
+		 * Barrier clones share an original request.
+		 * Leave it to dm_end_request(), which handles this special
+		 * case.
+		 */
+		BUG_ON(error > 0);
+		dm_end_request(clone, error);
+		return;
+	}
+
 	rq->cmd_flags |= REQ_FAILED;
 	dm_complete_request(clone, error);
 }
@@ -1214,7 +1299,7 @@ static void __split_and_process_bio(struct mapped_device *md, struct bio *bio)
 	struct clone_info ci;
 	int error = 0;
 
-	ci.map = dm_get_table(md);
+	ci.map = dm_get_live_table(md);
 	if (unlikely(!ci.map)) {
 		if (!bio_rw_flagged(bio, BIO_RW_BARRIER))
 			bio_io_error(bio);
@@ -1255,7 +1340,7 @@ static int dm_merge_bvec(struct request_queue *q,
 			 struct bio_vec *biovec)
 {
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table(md);
 	struct dm_target *ti;
 	sector_t max_sectors;
 	int max_size = 0;
@@ -1352,11 +1437,6 @@ static int dm_make_request(struct request_queue *q, struct bio *bio)
 {
 	struct mapped_device *md = q->queuedata;
 
-	if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
-	}
-
 	return md->saved_make_request_fn(q, bio); /* call __make_request() */
 }
 
@@ -1375,6 +1455,25 @@ static int dm_request(struct request_queue *q, struct bio *bio)
 	return _dm_request(q, bio);
 }
 
+/*
+ * Mark this request as flush request, so that dm_request_fn() can
+ * recognize.
+ */
+static void dm_rq_prepare_flush(struct request_queue *q, struct request *rq)
+{
+	rq->cmd_type = REQ_TYPE_LINUX_BLOCK;
+	rq->cmd[0] = REQ_LB_OP_FLUSH;
+}
+
+static bool dm_rq_is_flush_request(struct request *rq)
+{
+	if (rq->cmd_type == REQ_TYPE_LINUX_BLOCK &&
+	    rq->cmd[0] == REQ_LB_OP_FLUSH)
+		return true;
+	else
+		return false;
+}
+
 void dm_dispatch_request(struct request *rq)
 {
 	int r;
@@ -1420,25 +1519,54 @@ static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
 static int setup_clone(struct request *clone, struct request *rq,
 		       struct dm_rq_target_io *tio)
 {
-	int r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
-				  dm_rq_bio_constructor, tio);
+	int r;
 
-	if (r)
-		return r;
+	if (dm_rq_is_flush_request(rq)) {
+		blk_rq_init(NULL, clone);
+		clone->cmd_type = REQ_TYPE_FS;
+		clone->cmd_flags |= (REQ_HARDBARRIER | WRITE);
+	} else {
+		r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
+				      dm_rq_bio_constructor, tio);
+		if (r)
+			return r;
+
+		clone->cmd = rq->cmd;
+		clone->cmd_len = rq->cmd_len;
+		clone->sense = rq->sense;
+		clone->buffer = rq->buffer;
+	}
 
-	clone->cmd = rq->cmd;
-	clone->cmd_len = rq->cmd_len;
-	clone->sense = rq->sense;
-	clone->buffer = rq->buffer;
 	clone->end_io = end_clone_request;
 	clone->end_io_data = tio;
 
 	return 0;
 }
 
-static int dm_rq_flush_suspending(struct mapped_device *md)
+static struct request *clone_rq(struct request *rq, struct mapped_device *md,
+				gfp_t gfp_mask)
 {
-	return !md->suspend_rq.special;
+	struct request *clone;
+	struct dm_rq_target_io *tio;
+
+	tio = alloc_rq_tio(md, gfp_mask);
+	if (!tio)
+		return NULL;
+
+	tio->md = md;
+	tio->ti = NULL;
+	tio->orig = rq;
+	tio->error = 0;
+	memset(&tio->info, 0, sizeof(tio->info));
+
+	clone = &tio->clone;
+	if (setup_clone(clone, rq, tio)) {
+		/* -ENOMEM */
+		free_rq_tio(tio);
+		return NULL;
+	}
+
+	return clone;
 }
 
 /*
@@ -1447,51 +1575,30 @@ static int dm_rq_flush_suspending(struct mapped_device *md)
 static int dm_prep_fn(struct request_queue *q, struct request *rq)
 {
 	struct mapped_device *md = q->queuedata;
-	struct dm_rq_target_io *tio;
 	struct request *clone;
 
-	if (unlikely(rq == &md->suspend_rq)) {
-		if (dm_rq_flush_suspending(md))
-			return BLKPREP_OK;
-		else
-			/* The flush suspend was interrupted */
-			return BLKPREP_KILL;
-	}
+	if (unlikely(dm_rq_is_flush_request(rq)))
+		return BLKPREP_OK;
 
 	if (unlikely(rq->special)) {
 		DMWARN("Already has something in rq->special.");
 		return BLKPREP_KILL;
 	}
 
-	tio = alloc_rq_tio(md); /* Only one for each original request */
-	if (!tio)
-		/* -ENOMEM */
+	clone = clone_rq(rq, md, GFP_ATOMIC);
+	if (!clone)
 		return BLKPREP_DEFER;
 
-	tio->md = md;
-	tio->ti = NULL;
-	tio->orig = rq;
-	tio->error = 0;
-	memset(&tio->info, 0, sizeof(tio->info));
-
-	clone = &tio->clone;
-	if (setup_clone(clone, rq, tio)) {
-		/* -ENOMEM */
-		free_rq_tio(tio);
-		return BLKPREP_DEFER;
-	}
-
 	rq->special = clone;
 	rq->cmd_flags |= REQ_DONTPREP;
 
 	return BLKPREP_OK;
 }
 
-static void map_request(struct dm_target *ti, struct request *rq,
+static void map_request(struct dm_target *ti, struct request *clone,
 			struct mapped_device *md)
 {
 	int r;
-	struct request *clone = rq->special;
 	struct dm_rq_target_io *tio = clone->end_io_data;
 
 	/*
@@ -1536,29 +1643,26 @@ static void map_request(struct dm_target *ti, struct request *rq,
 static void dm_request_fn(struct request_queue *q)
 {
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table(md);
 	struct dm_target *ti;
-	struct request *rq;
+	struct request *rq, *clone;
 
 	/*
-	 * For noflush suspend, check blk_queue_stopped() to immediately
-	 * quit I/O dispatching.
+	 * For suspend, check blk_queue_stopped() and increment
+	 * ->pending within a single queue_lock not to increment the
+	 * number of in-flight I/Os after the queue is stopped in
+	 * dm_suspend().
 	 */
 	while (!blk_queue_plugged(q) && !blk_queue_stopped(q)) {
 		rq = blk_peek_request(q);
 		if (!rq)
 			goto plug_and_out;
 
-		if (unlikely(rq == &md->suspend_rq)) { /* Flush suspend maker */
-			if (queue_in_flight(q))
-				/* Not quiet yet.  Wait more */
-				goto plug_and_out;
-
-			/* This device should be quiet now */
-			__stop_queue(q);
+		if (unlikely(dm_rq_is_flush_request(rq))) {
+			BUG_ON(md->flush_request);
+			md->flush_request = rq;
 			blk_start_request(rq);
-			__blk_end_request_all(rq, 0);
-			wake_up(&md->wait);
+			queue_work(md->wq, &md->barrier_work);
 			goto out;
 		}
 
@@ -1567,8 +1671,11 @@ static void dm_request_fn(struct request_queue *q)
 			goto plug_and_out;
 
 		blk_start_request(rq);
+		clone = rq->special;
+		atomic_inc(&md->pending[rq_data_dir(clone)]);
+
 		spin_unlock(q->queue_lock);
-		map_request(ti, rq, md);
+		map_request(ti, clone, md);
 		spin_lock_irq(q->queue_lock);
 	}
 
@@ -1595,7 +1702,7 @@ static int dm_lld_busy(struct request_queue *q)
 {
 	int r;
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table(md);
 
 	if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))
 		r = 1;
@@ -1610,7 +1717,7 @@ static int dm_lld_busy(struct request_queue *q)
 static void dm_unplug_all(struct request_queue *q)
 {
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table(md);
 
 	if (map) {
 		if (dm_request_based(md))
@@ -1628,7 +1735,7 @@ static int dm_any_congested(void *congested_data, int bdi_bits)
 	struct dm_table *map;
 
 	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
-		map = dm_get_table(md);
+		map = dm_get_live_table(md);
 		if (map) {
 			/*
 			 * Request-based dm cares about only own queue for
@@ -1725,6 +1832,7 @@ out:
 static const struct block_device_operations dm_blk_dops;
 
 static void dm_wq_work(struct work_struct *work);
+static void dm_rq_barrier_work(struct work_struct *work);
 
 /*
  * Allocate and initialise a blank device with a given minor.
@@ -1754,6 +1862,7 @@ static struct mapped_device *alloc_dev(int minor)
 	init_rwsem(&md->io_lock);
 	mutex_init(&md->suspend_lock);
 	spin_lock_init(&md->deferred_lock);
+	spin_lock_init(&md->barrier_error_lock);
 	rwlock_init(&md->map_lock);
 	atomic_set(&md->holders, 1);
 	atomic_set(&md->open_count, 0);
@@ -1788,6 +1897,8 @@ static struct mapped_device *alloc_dev(int minor)
 	blk_queue_softirq_done(md->queue, dm_softirq_done);
 	blk_queue_prep_rq(md->queue, dm_prep_fn);
 	blk_queue_lld_busy(md->queue, dm_lld_busy);
+	blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN_FLUSH,
+			  dm_rq_prepare_flush);
 
 	md->disk = alloc_disk(1);
 	if (!md->disk)
@@ -1797,6 +1908,7 @@ static struct mapped_device *alloc_dev(int minor)
 	atomic_set(&md->pending[1], 0);
 	init_waitqueue_head(&md->wait);
 	INIT_WORK(&md->work, dm_wq_work);
+	INIT_WORK(&md->barrier_work, dm_rq_barrier_work);
 	init_waitqueue_head(&md->eventq);
 
 	md->disk->major = _major;
@@ -1921,9 +2033,13 @@ static void __set_size(struct mapped_device *md, sector_t size)
 	mutex_unlock(&md->bdev->bd_inode->i_mutex);
 }
 
-static int __bind(struct mapped_device *md, struct dm_table *t,
-		  struct queue_limits *limits)
+/*
+ * Returns old map, which caller must destroy.
+ */
+static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
+			       struct queue_limits *limits)
 {
+	struct dm_table *old_map;
 	struct request_queue *q = md->queue;
 	sector_t size;
 	unsigned long flags;
@@ -1938,11 +2054,6 @@ static int __bind(struct mapped_device *md, struct dm_table *t,
 
 	__set_size(md, size);
 
-	if (!size) {
-		dm_table_destroy(t);
-		return 0;
-	}
-
 	dm_table_event_callback(t, event_callback, md);
 
 	/*
@@ -1958,26 +2069,31 @@ static int __bind(struct mapped_device *md, struct dm_table *t,
 	__bind_mempools(md, t);
 
 	write_lock_irqsave(&md->map_lock, flags);
+	old_map = md->map;
 	md->map = t;
 	dm_table_set_restrictions(t, q, limits);
 	write_unlock_irqrestore(&md->map_lock, flags);
 
-	return 0;
+	return old_map;
 }
 
-static void __unbind(struct mapped_device *md)
+/*
+ * Returns unbound table for the caller to free.
+ */
+static struct dm_table *__unbind(struct mapped_device *md)
 {
 	struct dm_table *map = md->map;
 	unsigned long flags;
 
 	if (!map)
-		return;
+		return NULL;
 
 	dm_table_event_callback(map, NULL, NULL);
 	write_lock_irqsave(&md->map_lock, flags);
 	md->map = NULL;
 	write_unlock_irqrestore(&md->map_lock, flags);
-	dm_table_destroy(map);
+
+	return map;
 }
 
 /*
@@ -2059,7 +2175,7 @@ void dm_put(struct mapped_device *md)
 	BUG_ON(test_bit(DMF_FREEING, &md->flags));
 
 	if (atomic_dec_and_lock(&md->holders, &_minor_lock)) {
-		map = dm_get_table(md);
+		map = dm_get_live_table(md);
 		idr_replace(&_minor_idr, MINOR_ALLOCED,
 			    MINOR(disk_devt(dm_disk(md))));
 		set_bit(DMF_FREEING, &md->flags);
@@ -2070,7 +2186,7 @@ void dm_put(struct mapped_device *md)
 		}
 		dm_sysfs_exit(md);
 		dm_table_put(map);
-		__unbind(md);
+		dm_table_destroy(__unbind(md));
 		free_dev(md);
 	}
 }
@@ -2080,8 +2196,6 @@ static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
 {
 	int r = 0;
 	DECLARE_WAITQUEUE(wait, current);
-	struct request_queue *q = md->queue;
-	unsigned long flags;
 
 	dm_unplug_all(md->queue);
 
@@ -2091,15 +2205,7 @@ static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
 		set_current_state(interruptible);
 
 		smp_mb();
-		if (dm_request_based(md)) {
-			spin_lock_irqsave(q->queue_lock, flags);
-			if (!queue_in_flight(q) && blk_queue_stopped(q)) {
-				spin_unlock_irqrestore(q->queue_lock, flags);
-				break;
-			}
-			spin_unlock_irqrestore(q->queue_lock, flags);
-		} else if (!atomic_read(&md->pending[0]) &&
-					!atomic_read(&md->pending[1]))
+		if (!md_in_flight(md))
 			break;
 
 		if (interruptible == TASK_INTERRUPTIBLE &&
@@ -2194,13 +2300,81 @@ static void dm_queue_flush(struct mapped_device *md)
 	queue_work(md->wq, &md->work);
 }
 
+static void dm_rq_set_flush_nr(struct request *clone, unsigned flush_nr)
+{
+	struct dm_rq_target_io *tio = clone->end_io_data;
+
+	tio->info.flush_request = flush_nr;
+}
+
+/* Issue barrier requests to targets and wait for their completion. */
+static int dm_rq_barrier(struct mapped_device *md)
+{
+	int i, j;
+	struct dm_table *map = dm_get_live_table(md);
+	unsigned num_targets = dm_table_get_num_targets(map);
+	struct dm_target *ti;
+	struct request *clone;
+
+	md->barrier_error = 0;
+
+	for (i = 0; i < num_targets; i++) {
+		ti = dm_table_get_target(map, i);
+		for (j = 0; j < ti->num_flush_requests; j++) {
+			clone = clone_rq(md->flush_request, md, GFP_NOIO);
+			dm_rq_set_flush_nr(clone, j);
+			atomic_inc(&md->pending[rq_data_dir(clone)]);
+			map_request(ti, clone, md);
+		}
+	}
+
+	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
+	dm_table_put(map);
+
+	return md->barrier_error;
+}
+
+static void dm_rq_barrier_work(struct work_struct *work)
+{
+	int error;
+	struct mapped_device *md = container_of(work, struct mapped_device,
+						barrier_work);
+	struct request_queue *q = md->queue;
+	struct request *rq;
+	unsigned long flags;
+
+	/*
+	 * Hold the md reference here and leave it at the last part so that
+	 * the md can't be deleted by device opener when the barrier request
+	 * completes.
+	 */
+	dm_get(md);
+
+	error = dm_rq_barrier(md);
+
+	rq = md->flush_request;
+	md->flush_request = NULL;
+
+	if (error == DM_ENDIO_REQUEUE) {
+		spin_lock_irqsave(q->queue_lock, flags);
+		blk_requeue_request(q, rq);
+		spin_unlock_irqrestore(q->queue_lock, flags);
+	} else
+		blk_end_request_all(rq, error);
+
+	blk_run_queue(q);
+
+	dm_put(md);
+}
+
 /*
- * Swap in a new table (destroying old one).
+ * Swap in a new table, returning the old one for the caller to destroy.
  */
-int dm_swap_table(struct mapped_device *md, struct dm_table *table)
+struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
 {
+	struct dm_table *map = ERR_PTR(-EINVAL);
 	struct queue_limits limits;
-	int r = -EINVAL;
+	int r;
 
 	mutex_lock(&md->suspend_lock);
 
@@ -2209,8 +2383,10 @@ int dm_swap_table(struct mapped_device *md, struct dm_table *table)
 		goto out;
 
 	r = dm_calculate_queue_limits(table, &limits);
-	if (r)
+	if (r) {
+		map = ERR_PTR(r);
 		goto out;
+	}
 
 	/* cannot change the device type, once a table is bound */
 	if (md->map &&
@@ -2219,73 +2395,11 @@ int dm_swap_table(struct mapped_device *md, struct dm_table *table)
 		goto out;
 	}
 
-	__unbind(md);
-	r = __bind(md, table, &limits);
+	map = __bind(md, table, &limits);
 
 out:
 	mutex_unlock(&md->suspend_lock);
-	return r;
-}
-
-static void dm_rq_invalidate_suspend_marker(struct mapped_device *md)
-{
-	md->suspend_rq.special = (void *)0x1;
-}
-
-static void dm_rq_abort_suspend(struct mapped_device *md, int noflush)
-{
-	struct request_queue *q = md->queue;
-	unsigned long flags;
-
-	spin_lock_irqsave(q->queue_lock, flags);
-	if (!noflush)
-		dm_rq_invalidate_suspend_marker(md);
-	__start_queue(q);
-	spin_unlock_irqrestore(q->queue_lock, flags);
-}
-
-static void dm_rq_start_suspend(struct mapped_device *md, int noflush)
-{
-	struct request *rq = &md->suspend_rq;
-	struct request_queue *q = md->queue;
-
-	if (noflush)
-		stop_queue(q);
-	else {
-		blk_rq_init(q, rq);
-		blk_insert_request(q, rq, 0, NULL);
-	}
-}
-
-static int dm_rq_suspend_available(struct mapped_device *md, int noflush)
-{
-	int r = 1;
-	struct request *rq = &md->suspend_rq;
-	struct request_queue *q = md->queue;
-	unsigned long flags;
-
-	if (noflush)
-		return r;
-
-	/* The marker must be protected by queue lock if it is in use */
-	spin_lock_irqsave(q->queue_lock, flags);
-	if (unlikely(rq->ref_count)) {
-		/*
-		 * This can happen, when the previous flush suspend was
-		 * interrupted, the marker is still in the queue and
-		 * this flush suspend has been invoked, because we don't
-		 * remove the marker at the time of suspend interruption.
-		 * We have only one marker per mapped_device, so we can't
-		 * start another flush suspend while it is in use.
-		 */
-		BUG_ON(!rq->special); /* The marker should be invalidated */
-		DMWARN("Invalidating the previous flush suspend is still in"
-		       " progress.  Please retry later.");
-		r = 0;
-	}
-	spin_unlock_irqrestore(q->queue_lock, flags);
-
-	return r;
+	return map;
 }
 
 /*
@@ -2330,49 +2444,11 @@ static void unlock_fs(struct mapped_device *md)
 /*
  * Suspend mechanism in request-based dm.
  *
- * After the suspend starts, further incoming requests are kept in
- * the request_queue and deferred.
- * Remaining requests in the request_queue at the start of suspend are flushed
- * if it is flush suspend.
- * The suspend completes when the following conditions have been satisfied,
- * so wait for it:
- *    1. q->in_flight is 0 (which means no in_flight request)
- *    2. queue has been stopped (which means no request dispatching)
- *
- *
- * Noflush suspend
- * ---------------
- * Noflush suspend doesn't need to dispatch remaining requests.
- * So stop the queue immediately.  Then, wait for all in_flight requests
- * to be completed or requeued.
- *
- * To abort noflush suspend, start the queue.
- *
+ * 1. Flush all I/Os by lock_fs() if needed.
+ * 2. Stop dispatching any I/O by stopping the request_queue.
+ * 3. Wait for all in-flight I/Os to be completed or requeued.
  *
- * Flush suspend
- * -------------
- * Flush suspend needs to dispatch remaining requests.  So stop the queue
- * after the remaining requests are completed. (Requeued request must be also
- * re-dispatched and completed.  Until then, we can't stop the queue.)
- *
- * During flushing the remaining requests, further incoming requests are also
- * inserted to the same queue.  To distinguish which requests are to be
- * flushed, we insert a marker request to the queue at the time of starting
- * flush suspend, like a barrier.
- * The dispatching is blocked when the marker is found on the top of the queue.
- * And the queue is stopped when all in_flight requests are completed, since
- * that means the remaining requests are completely flushed.
- * Then, the marker is removed from the queue.
- *
- * To abort flush suspend, we also need to take care of the marker, not only
- * starting the queue.
- * We don't remove the marker forcibly from the queue since it's against
- * the block-layer manner.  Instead, we put a invalidated mark on the marker.
- * When the invalidated marker is found on the top of the queue, it is
- * immediately removed from the queue, so it doesn't block dispatching.
- * Because we have only one marker per mapped_device, we can't start another
- * flush suspend until the invalidated marker is removed from the queue.
- * So fail and return with -EBUSY in such a case.
+ * To abort suspend, start the request_queue.
  */
 int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 {
@@ -2388,12 +2464,7 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 		goto out_unlock;
 	}
 
-	if (dm_request_based(md) && !dm_rq_suspend_available(md, noflush)) {
-		r = -EBUSY;
-		goto out_unlock;
-	}
-
-	map = dm_get_table(md);
+	map = dm_get_live_table(md);
 
 	/*
 	 * DMF_NOFLUSH_SUSPENDING must be set before presuspend.
@@ -2406,8 +2477,10 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 	dm_table_presuspend_targets(map);
 
 	/*
-	 * Flush I/O to the device. noflush supersedes do_lockfs,
-	 * because lock_fs() needs to flush I/Os.
+	 * Flush I/O to the device.
+	 * Any I/O submitted after lock_fs() may not be flushed.
+	 * noflush takes precedence over do_lockfs.
+	 * (lock_fs() flushes I/Os and waits for them to complete.)
 	 */
 	if (!noflush && do_lockfs) {
 		r = lock_fs(md);
@@ -2436,10 +2509,15 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 	set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags);
 	up_write(&md->io_lock);
 
-	flush_workqueue(md->wq);
-
+	/*
+	 * Request-based dm uses md->wq for barrier (dm_rq_barrier_work) which
+	 * can be kicked until md->queue is stopped.  So stop md->queue before
+	 * flushing md->wq.
+	 */
 	if (dm_request_based(md))
-		dm_rq_start_suspend(md, noflush);
+		stop_queue(md->queue);
+
+	flush_workqueue(md->wq);
 
 	/*
 	 * At this point no more requests are entering target request routines.
@@ -2458,7 +2536,7 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 		dm_queue_flush(md);
 
 		if (dm_request_based(md))
-			dm_rq_abort_suspend(md, noflush);
+			start_queue(md->queue);
 
 		unlock_fs(md);
 		goto out; /* pushback list is already flushed, so skip flush */
@@ -2491,7 +2569,7 @@ int dm_resume(struct mapped_device *md)
 	if (!dm_suspended(md))
 		goto out;
 
-	map = dm_get_table(md);
+	map = dm_get_live_table(md);
 	if (!map || !dm_table_get_size(map))
 		goto out;
 
@@ -2592,7 +2670,7 @@ struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
 		return NULL;
 
 	if (test_bit(DMF_FREEING, &md->flags) ||
-	    test_bit(DMF_DELETING, &md->flags))
+	    dm_deleting_md(md))
 		return NULL;
 
 	dm_get(md);
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index a7663eba17e2..604a5f2a2383 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -89,6 +89,11 @@ int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 int dm_split_args(int *argc, char ***argvp, char *input);
 
 /*
+ * Is this mapped_device being deleted?
+ */
+int dm_deleting_md(struct mapped_device *md);
+
+/*
  * The device-mapper can be driven through one of two interfaces;
  * ioctl or filesystem, depending which patch you have applied.
  */
@@ -118,6 +123,9 @@ int dm_lock_for_deletion(struct mapped_device *md);
 void dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
 		       unsigned cookie);
 
+int dm_io_init(void);
+void dm_io_exit(void);
+
 int dm_kcopyd_init(void);
 void dm_kcopyd_exit(void);
 
diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h
index df7607e6dce8..b9c6c8ca11be 100644
--- a/include/linux/device-mapper.h
+++ b/include/linux/device-mapper.h
@@ -276,7 +276,7 @@ void dm_table_unplug_all(struct dm_table *t);
 /*
  * Table reference counting.
  */
-struct dm_table *dm_get_table(struct mapped_device *md);
+struct dm_table *dm_get_live_table(struct mapped_device *md);
 void dm_table_get(struct dm_table *t);
 void dm_table_put(struct dm_table *t);
 
@@ -295,8 +295,10 @@ void dm_table_event(struct dm_table *t);
 
 /*
  * The device must be suspended before calling this method.
+ * Returns the previous table, which the caller must destroy.
  */
-int dm_swap_table(struct mapped_device *md, struct dm_table *t);
+struct dm_table *dm_swap_table(struct mapped_device *md,
+			       struct dm_table *t);
 
 /*
  * A wrapper around vmalloc.
diff --git a/include/linux/dm-ioctl.h b/include/linux/dm-ioctl.h
index 2ab84c83c31a..aa95508d2f95 100644
--- a/include/linux/dm-ioctl.h
+++ b/include/linux/dm-ioctl.h
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001 - 2003 Sistina Software (UK) Limited.
- * Copyright (C) 2004 - 2005 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004 - 2009 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the LGPL.
  */
@@ -266,9 +266,9 @@ enum {
 #define DM_DEV_SET_GEOMETRY	_IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl)
 
 #define DM_VERSION_MAJOR	4
-#define DM_VERSION_MINOR	15
+#define DM_VERSION_MINOR	16
 #define DM_VERSION_PATCHLEVEL	0
-#define DM_VERSION_EXTRA	"-ioctl (2009-04-01)"
+#define DM_VERSION_EXTRA	"-ioctl (2009-11-05)"
 
 /* Status bits */
 #define DM_READONLY_FLAG	(1 << 0) /* In/Out */
@@ -309,4 +309,11 @@ enum {
  */
 #define DM_NOFLUSH_FLAG		(1 << 11) /* In */
 
+/*
+ * If set, any table information returned will relate to the inactive
+ * table instead of the live one.  Always check DM_INACTIVE_PRESENT_FLAG
+ * is set before using the data returned.
+ */
+#define DM_QUERY_INACTIVE_TABLE_FLAG	(1 << 12) /* In */
+
 #endif				/* _LINUX_DM_IOCTL_H */