Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm: (80 commits)
  dm snapshot: use merge origin if snapshot invalid
  dm snapshot: report merge failure in status
  dm snapshot: merge consecutive chunks together
  dm snapshot: trigger exceptions in remaining snapshots during merge
  dm snapshot: delay merging a chunk until writes to it complete
  dm snapshot: queue writes to chunks being merged
  dm snapshot: add merging
  dm snapshot: permit only one merge at once
  dm snapshot: support barriers in snapshot merge target
  dm snapshot: avoid allocating exceptions in merge
  dm snapshot: rework writing to origin
  dm snapshot: add merge target
  dm exception store: add merge specific methods
  dm snapshot: create function for chunk_is_tracked wait
  dm snapshot: make bio optional in __origin_write
  dm mpath: reject messages when device is suspended
  dm: export suspended state to targets
  dm: rename dm_suspended to dm_suspended_md
  dm: swap target postsuspend call and setting suspended flag
  dm crypt: add plain64 iv
  ...

23 files changed, 2349 insertions, 889 deletions

diff --git a/Documentation/device-mapper/snapshot.txt b/Documentation/device-mapper/snapshot.txt
index a5009c8300f..e3a77b21513 100644
--- a/Documentation/device-mapper/snapshot.txt
+++ b/Documentation/device-mapper/snapshot.txt
@@ -8,13 +8,19 @@ the block device which are also writable without interfering with the
 original content;
 *) To create device "forks", i.e. multiple different versions of the
 same data stream.
+*) To merge a snapshot of a block device back into the snapshot's origin
+device.
 
+In the first two cases, dm copies only the chunks of data that get
+changed and uses a separate copy-on-write (COW) block device for
+storage.
 
-In both cases, dm copies only the chunks of data that get changed and
-uses a separate copy-on-write (COW) block device for storage.
+For snapshot merge the contents of the COW storage are merged back into
+the origin device.
 
 
-There are two dm targets available: snapshot and snapshot-origin.
+There are three dm targets available:
+snapshot, snapshot-origin, and snapshot-merge.
 
 *) snapshot-origin <origin>
 
@@ -40,8 +46,25 @@ The difference is that for transient snapshots less metadata must be
 saved on disk - they can be kept in memory by the kernel.
 
 
-How this is used by LVM2
-========================
+* snapshot-merge <origin> <COW device> <persistent> <chunksize>
+
+takes the same table arguments as the snapshot target except it only
+works with persistent snapshots.  This target assumes the role of the
+"snapshot-origin" target and must not be loaded if the "snapshot-origin"
+is still present for <origin>.
+
+Creates a merging snapshot that takes control of the changed chunks
+stored in the <COW device> of an existing snapshot, through a handover
+procedure, and merges these chunks back into the <origin>.  Once merging
+has started (in the background) the <origin> may be opened and the merge
+will continue while I/O is flowing to it.  Changes to the <origin> are
+deferred until the merging snapshot's corresponding chunk(s) have been
+merged.  Once merging has started the snapshot device, associated with
+the "snapshot" target, will return -EIO when accessed.
+
+
+How snapshot is used by LVM2
+============================
 When you create the first LVM2 snapshot of a volume, four dm devices are used:
 
 1) a device containing the original mapping table of the source volume;
@@ -72,3 +95,30 @@ brw-------  1 root root 254, 12 29 ago 18:15 /dev/mapper/volumeGroup-snap-cow
 brw-------  1 root root 254, 13 29 ago 18:15 /dev/mapper/volumeGroup-snap
 brw-------  1 root root 254, 10 29 ago 18:14 /dev/mapper/volumeGroup-base
 
+
+How snapshot-merge is used by LVM2
+==================================
+A merging snapshot assumes the role of the "snapshot-origin" while
+merging.  As such the "snapshot-origin" is replaced with
+"snapshot-merge".  The "-real" device is not changed and the "-cow"
+device is renamed to <origin name>-cow to aid LVM2's cleanup of the
+merging snapshot after it completes.  The "snapshot" that hands over its
+COW device to the "snapshot-merge" is deactivated (unless using lvchange
+--refresh); but if it is left active it will simply return I/O errors.
+
+A snapshot will merge into its origin with the following command:
+
+lvconvert --merge volumeGroup/snap
+
+we'll now have this situation:
+
+# dmsetup table|grep volumeGroup
+
+volumeGroup-base-real: 0 2097152 linear 8:19 384
+volumeGroup-base-cow: 0 204800 linear 8:19 2097536
+volumeGroup-base: 0 2097152 snapshot-merge 254:11 254:12 P 16
+
+# ls -lL /dev/mapper/volumeGroup-*
+brw-------  1 root root 254, 11 29 ago 18:15 /dev/mapper/volumeGroup-base-real
+brw-------  1 root root 254, 12 29 ago 18:16 /dev/mapper/volumeGroup-base-cow
+brw-------  1 root root 254, 10 29 ago 18:16 /dev/mapper/volumeGroup-base
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index e412980763b..a93637223c8 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1,7 +1,7 @@
 /*
  * Copyright (C) 2003 Christophe Saout <christophe@saout.de>
  * Copyright (C) 2004 Clemens Fruhwirth <clemens@endorphin.org>
- * Copyright (C) 2006-2008 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2006-2009 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
@@ -71,10 +71,21 @@ struct crypt_iv_operations {
 	int (*ctr)(struct crypt_config *cc, struct dm_target *ti,
 		   const char *opts);
 	void (*dtr)(struct crypt_config *cc);
-	const char *(*status)(struct crypt_config *cc);
+	int (*init)(struct crypt_config *cc);
+	int (*wipe)(struct crypt_config *cc);
 	int (*generator)(struct crypt_config *cc, u8 *iv, sector_t sector);
 };
 
+struct iv_essiv_private {
+	struct crypto_cipher *tfm;
+	struct crypto_hash *hash_tfm;
+	u8 *salt;
+};
+
+struct iv_benbi_private {
+	int shift;
+};
+
 /*
  * Crypt: maps a linear range of a block device
  * and encrypts / decrypts at the same time.
@@ -102,8 +113,8 @@ struct crypt_config {
 	struct crypt_iv_operations *iv_gen_ops;
 	char *iv_mode;
 	union {
-		struct crypto_cipher *essiv_tfm;
-		int benbi_shift;
+		struct iv_essiv_private essiv;
+		struct iv_benbi_private benbi;
 	} iv_gen_private;
 	sector_t iv_offset;
 	unsigned int iv_size;
@@ -147,6 +158,9 @@ static void kcryptd_queue_crypt(struct dm_crypt_io *io);
  * plain: the initial vector is the 32-bit little-endian version of the sector
  *        number, padded with zeros if necessary.
  *
+ * plain64: the initial vector is the 64-bit little-endian version of the sector
+ *        number, padded with zeros if necessary.
+ *
  * essiv: "encrypted sector|salt initial vector", the sector number is
  *        encrypted with the bulk cipher using a salt as key. The salt
  *        should be derived from the bulk cipher's key via hashing.
@@ -169,88 +183,123 @@ static int crypt_iv_plain_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
 	return 0;
 }
 
-static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
-			      const char *opts)
+static int crypt_iv_plain64_gen(struct crypt_config *cc, u8 *iv,
+				sector_t sector)
 {
-	struct crypto_cipher *essiv_tfm;
-	struct crypto_hash *hash_tfm;
+	memset(iv, 0, cc->iv_size);
+	*(u64 *)iv = cpu_to_le64(sector);
+
+	return 0;
+}
+
+/* Initialise ESSIV - compute salt but no local memory allocations */
+static int crypt_iv_essiv_init(struct crypt_config *cc)
+{
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
 	struct hash_desc desc;
 	struct scatterlist sg;
-	unsigned int saltsize;
-	u8 *salt;
 	int err;
 
-	if (opts == NULL) {
+	sg_init_one(&sg, cc->key, cc->key_size);
+	desc.tfm = essiv->hash_tfm;
+	desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+
+	err = crypto_hash_digest(&desc, &sg, cc->key_size, essiv->salt);
+	if (err)
+		return err;
+
+	return crypto_cipher_setkey(essiv->tfm, essiv->salt,
+				    crypto_hash_digestsize(essiv->hash_tfm));
+}
+
+/* Wipe salt and reset key derived from volume key */
+static int crypt_iv_essiv_wipe(struct crypt_config *cc)
+{
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
+	unsigned salt_size = crypto_hash_digestsize(essiv->hash_tfm);
+
+	memset(essiv->salt, 0, salt_size);
+
+	return crypto_cipher_setkey(essiv->tfm, essiv->salt, salt_size);
+}
+
+static void crypt_iv_essiv_dtr(struct crypt_config *cc)
+{
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
+
+	crypto_free_cipher(essiv->tfm);
+	essiv->tfm = NULL;
+
+	crypto_free_hash(essiv->hash_tfm);
+	essiv->hash_tfm = NULL;
+
+	kzfree(essiv->salt);
+	essiv->salt = NULL;
+}
+
+static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
+			      const char *opts)
+{
+	struct crypto_cipher *essiv_tfm = NULL;
+	struct crypto_hash *hash_tfm = NULL;
+	u8 *salt = NULL;
+	int err;
+
+	if (!opts) {
 		ti->error = "Digest algorithm missing for ESSIV mode";
 		return -EINVAL;
 	}
 
-	/* Hash the cipher key with the given hash algorithm */
+	/* Allocate hash algorithm */
 	hash_tfm = crypto_alloc_hash(opts, 0, CRYPTO_ALG_ASYNC);
 	if (IS_ERR(hash_tfm)) {
 		ti->error = "Error initializing ESSIV hash";
-		return PTR_ERR(hash_tfm);
+		err = PTR_ERR(hash_tfm);
+		goto bad;
 	}
 
-	saltsize = crypto_hash_digestsize(hash_tfm);
-	salt = kmalloc(saltsize, GFP_KERNEL);
-	if (salt == NULL) {
+	salt = kzalloc(crypto_hash_digestsize(hash_tfm), GFP_KERNEL);
+	if (!salt) {
 		ti->error = "Error kmallocing salt storage in ESSIV";
-		crypto_free_hash(hash_tfm);
-		return -ENOMEM;
+		err = -ENOMEM;
+		goto bad;
 	}
 
-	sg_init_one(&sg, cc->key, cc->key_size);
-	desc.tfm = hash_tfm;
-	desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
-	err = crypto_hash_digest(&desc, &sg, cc->key_size, salt);
-	crypto_free_hash(hash_tfm);
-
-	if (err) {
-		ti->error = "Error calculating hash in ESSIV";
-		kfree(salt);
-		return err;
-	}
-
-	/* Setup the essiv_tfm with the given salt */
+	/* Allocate essiv_tfm */
 	essiv_tfm = crypto_alloc_cipher(cc->cipher, 0, CRYPTO_ALG_ASYNC);
 	if (IS_ERR(essiv_tfm)) {
 		ti->error = "Error allocating crypto tfm for ESSIV";
-		kfree(salt);
-		return PTR_ERR(essiv_tfm);
+		err = PTR_ERR(essiv_tfm);
+		goto bad;
 	}
 	if (crypto_cipher_blocksize(essiv_tfm) !=
 	    crypto_ablkcipher_ivsize(cc->tfm)) {
 		ti->error = "Block size of ESSIV cipher does "
 			    "not match IV size of block cipher";
-		crypto_free_cipher(essiv_tfm);
-		kfree(salt);
-		return -EINVAL;
+		err = -EINVAL;
+		goto bad;
 	}
-	err = crypto_cipher_setkey(essiv_tfm, salt, saltsize);
-	if (err) {
-		ti->error = "Failed to set key for ESSIV cipher";
-		crypto_free_cipher(essiv_tfm);
-		kfree(salt);
-		return err;
-	}
-	kfree(salt);
 
-	cc->iv_gen_private.essiv_tfm = essiv_tfm;
+	cc->iv_gen_private.essiv.salt = salt;
+	cc->iv_gen_private.essiv.tfm = essiv_tfm;
+	cc->iv_gen_private.essiv.hash_tfm = hash_tfm;
+
 	return 0;
-}
 
-static void crypt_iv_essiv_dtr(struct crypt_config *cc)
-{
-	crypto_free_cipher(cc->iv_gen_private.essiv_tfm);
-	cc->iv_gen_private.essiv_tfm = NULL;
+bad:
+	if (essiv_tfm && !IS_ERR(essiv_tfm))
+		crypto_free_cipher(essiv_tfm);
+	if (hash_tfm && !IS_ERR(hash_tfm))
+		crypto_free_hash(hash_tfm);
+	kfree(salt);
+	return err;
 }
 
 static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
 {
 	memset(iv, 0, cc->iv_size);
 	*(u64 *)iv = cpu_to_le64(sector);
-	crypto_cipher_encrypt_one(cc->iv_gen_private.essiv_tfm, iv, iv);
+	crypto_cipher_encrypt_one(cc->iv_gen_private.essiv.tfm, iv, iv);
 	return 0;
 }
 
@@ -273,7 +322,7 @@ static int crypt_iv_benbi_ctr(struct crypt_config *cc, struct dm_target *ti,
 		return -EINVAL;
 	}
 
-	cc->iv_gen_private.benbi_shift = 9 - log;
+	cc->iv_gen_private.benbi.shift = 9 - log;
 
 	return 0;
 }
@@ -288,7 +337,7 @@ static int crypt_iv_benbi_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
 
 	memset(iv, 0, cc->iv_size - sizeof(u64)); /* rest is cleared below */
 
-	val = cpu_to_be64(((u64)sector << cc->iv_gen_private.benbi_shift) + 1);
+	val = cpu_to_be64(((u64)sector << cc->iv_gen_private.benbi.shift) + 1);
 	put_unaligned(val, (__be64 *)(iv + cc->iv_size - sizeof(u64)));
 
 	return 0;
@@ -305,9 +354,15 @@ static struct crypt_iv_operations crypt_iv_plain_ops = {
 	.generator = crypt_iv_plain_gen
 };
 
+static struct crypt_iv_operations crypt_iv_plain64_ops = {
+	.generator = crypt_iv_plain64_gen
+};
+
 static struct crypt_iv_operations crypt_iv_essiv_ops = {
 	.ctr       = crypt_iv_essiv_ctr,
 	.dtr       = crypt_iv_essiv_dtr,
+	.init      = crypt_iv_essiv_init,
+	.wipe      = crypt_iv_essiv_wipe,
 	.generator = crypt_iv_essiv_gen
 };
 
@@ -934,14 +989,14 @@ static int crypt_set_key(struct crypt_config *cc, char *key)
 
 	set_bit(DM_CRYPT_KEY_VALID, &cc->flags);
 
-	return 0;
+	return crypto_ablkcipher_setkey(cc->tfm, cc->key, cc->key_size);
 }
 
 static int crypt_wipe_key(struct crypt_config *cc)
 {
 	clear_bit(DM_CRYPT_KEY_VALID, &cc->flags);
 	memset(&cc->key, 0, cc->key_size * sizeof(u8));
-	return 0;
+	return crypto_ablkcipher_setkey(cc->tfm, cc->key, cc->key_size);
 }
 
 /*
@@ -983,11 +1038,6 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -ENOMEM;
 	}
 
- 	if (crypt_set_key(cc, argv[1])) {
-		ti->error = "Error decoding key";
-		goto bad_cipher;
-	}
-
 	/* Compatibility mode for old dm-crypt cipher strings */
 	if (!chainmode || (strcmp(chainmode, "plain") == 0 && !ivmode)) {
 		chainmode = "cbc";
@@ -1015,6 +1065,11 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	strcpy(cc->chainmode, chainmode);
 	cc->tfm = tfm;
 
+	if (crypt_set_key(cc, argv[1]) < 0) {
+		ti->error = "Error decoding and setting key";
+		goto bad_ivmode;
+	}
+
 	/*
 	 * Choose ivmode. Valid modes: "plain", "essiv:<esshash>", "benbi".
 	 * See comments at iv code
@@ -1024,6 +1079,8 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		cc->iv_gen_ops = NULL;
 	else if (strcmp(ivmode, "plain") == 0)
 		cc->iv_gen_ops = &crypt_iv_plain_ops;
+	else if (strcmp(ivmode, "plain64") == 0)
+		cc->iv_gen_ops = &crypt_iv_plain64_ops;
 	else if (strcmp(ivmode, "essiv") == 0)
 		cc->iv_gen_ops = &crypt_iv_essiv_ops;
 	else if (strcmp(ivmode, "benbi") == 0)
@@ -1039,6 +1096,12 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	    cc->iv_gen_ops->ctr(cc, ti, ivopts) < 0)
 		goto bad_ivmode;
 
+	if (cc->iv_gen_ops && cc->iv_gen_ops->init &&
+	    cc->iv_gen_ops->init(cc) < 0) {
+		ti->error = "Error initialising IV";
+		goto bad_slab_pool;
+	}
+
 	cc->iv_size = crypto_ablkcipher_ivsize(tfm);
 	if (cc->iv_size)
 		/* at least a 64 bit sector number should fit in our buffer */
@@ -1085,11 +1148,6 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		goto bad_bs;
 	}
 
-	if (crypto_ablkcipher_setkey(tfm, cc->key, key_size) < 0) {
-		ti->error = "Error setting key";
-		goto bad_device;
-	}
-
 	if (sscanf(argv[2], "%llu", &tmpll) != 1) {
 		ti->error = "Invalid iv_offset sector";
 		goto bad_device;
@@ -1278,6 +1336,7 @@ static void crypt_resume(struct dm_target *ti)
 static int crypt_message(struct dm_target *ti, unsigned argc, char **argv)
 {
 	struct crypt_config *cc = ti->private;
+	int ret = -EINVAL;
 
 	if (argc < 2)
 		goto error;
@@ -1287,10 +1346,22 @@ static int crypt_message(struct dm_target *ti, unsigned argc, char **argv)
 			DMWARN("not suspended during key manipulation.");
 			return -EINVAL;
 		}
-		if (argc == 3 && !strnicmp(argv[1], MESG_STR("set")))
-			return crypt_set_key(cc, argv[2]);
-		if (argc == 2 && !strnicmp(argv[1], MESG_STR("wipe")))
+		if (argc == 3 && !strnicmp(argv[1], MESG_STR("set"))) {
+			ret = crypt_set_key(cc, argv[2]);
+			if (ret)
+				return ret;
+			if (cc->iv_gen_ops && cc->iv_gen_ops->init)
+				ret = cc->iv_gen_ops->init(cc);
+			return ret;
+		}
+		if (argc == 2 && !strnicmp(argv[1], MESG_STR("wipe"))) {
+			if (cc->iv_gen_ops && cc->iv_gen_ops->wipe) {
+				ret = cc->iv_gen_ops->wipe(cc);
+				if (ret)
+					return ret;
+			}
 			return crypt_wipe_key(cc);
+		}
 	}
 
 error:
diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index 7dbe652efb5..2b7907b6dd0 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 8a223a48802..e8dfa06af3b 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;
@@ -64,17 +64,34 @@ 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);
 
 	/*
+	 * Returns 0 if the exception store is empty.
+	 *
+	 * If there are exceptions still to be merged, sets
+	 * *last_old_chunk and *last_new_chunk to the most recent
+	 * still-to-be-merged chunk and returns the number of
+	 * consecutive previous ones.
+	 */
+	int (*prepare_merge) (struct dm_exception_store *store,
+			      chunk_t *last_old_chunk, chunk_t *last_new_chunk);
+
+	/*
+	 * Clear the last n exceptions.
+	 * nr_merged must be <= the value returned by prepare_merge.
+	 */
+	int (*commit_merge) (struct dm_exception_store *store, int nr_merged);
+
+	/*
 	 * The snapshot is invalid, note this in the metadata.
 	 */
 	void (*drop_snapshot) (struct dm_exception_store *store);
@@ -86,19 +103,19 @@ 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 +126,11 @@ 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)
@@ -120,18 +142,25 @@ 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);
 
 	BUG_ON(!dm_consecutive_chunk_count(e));
 }
 
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
+{
+	BUG_ON(!dm_consecutive_chunk_count(e));
+
+	e->new_chunk -= (1ULL << DM_CHUNK_NUMBER_BITS);
+}
+
 #  else
 #    define DM_CHUNK_CONSECUTIVE_BITS 0
 
@@ -140,12 +169,16 @@ 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)
+{
+}
+
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
 {
 }
 
@@ -162,7 +195,7 @@ 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 +206,7 @@ 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 3a2e6a2f8bd..10f457ca6af 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