Merge tag 'v3.6-rc2' into drm-intel-next

Backmerge Linux 3.6-rc2 to resolve a few funny conflicts before we put
even more madness on top:

- drivers/gpu/drm/i915/i915_irq.c: Just a spurious WARN removed in
  -fixes, that has been changed in a variable-rename in -next, too.

- drivers/gpu/drm/i915/intel_ringbuffer.c: -next remove scratch_addr
  (since all their users have been extracted in another fucntion),
  -fixes added another user for a hw workaroudn.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

41 files changed, 1999 insertions, 1638 deletions

diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
index 10f122a3a85..d949b781f6f 100644
--- a/drivers/md/Kconfig
+++ b/drivers/md/Kconfig
@@ -260,15 +260,6 @@ config DM_DEBUG_BLOCK_STACK_TRACING
 
 	  If unsure, say N.
 
-config DM_DEBUG_SPACE_MAPS
-	boolean "Extra validation for thin provisioning space maps"
-	depends on DM_THIN_PROVISIONING
-	---help---
-	  Enable this for messages that may help debug problems with the
-	  space maps used by thin provisioning.
-
-          If unsure, say N.
-
 config DM_MIRROR
        tristate "Mirror target"
        depends on BLK_DEV_DM
@@ -277,13 +268,14 @@ config DM_MIRROR
          needed for live data migration tools such as 'pvmove'.
 
 config DM_RAID
-       tristate "RAID 1/4/5/6 target"
+       tristate "RAID 1/4/5/6/10 target"
        depends on BLK_DEV_DM
        select MD_RAID1
+       select MD_RAID10
        select MD_RAID456
        select BLK_DEV_MD
        ---help---
-	 A dm target that supports RAID1, RAID4, RAID5 and RAID6 mappings
+	 A dm target that supports RAID1, RAID10, RAID4, RAID5 and RAID6 mappings
 
 	 A RAID-5 set of N drives with a capacity of C MB per drive provides
 	 the capacity of C * (N - 1) MB, and protects against a failure
diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c
index 15dbe03117e..94e7f6ba2e1 100644
--- a/drivers/md/bitmap.c
+++ b/drivers/md/bitmap.c
@@ -1305,7 +1305,7 @@ int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sect
 			prepare_to_wait(&bitmap->overflow_wait, &__wait,
 					TASK_UNINTERRUPTIBLE);
 			spin_unlock_irq(&bitmap->counts.lock);
-			io_schedule();
+			schedule();
 			finish_wait(&bitmap->overflow_wait, &__wait);
 			continue;
 		}
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 3f06df59fd8..664743d6a6c 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -42,21 +42,21 @@ struct convert_context {
 	unsigned int offset_out;
 	unsigned int idx_in;
 	unsigned int idx_out;
-	sector_t sector;
-	atomic_t pending;
+	sector_t cc_sector;
+	atomic_t cc_pending;
 };
 
 /*
  * per bio private data
  */
 struct dm_crypt_io {
-	struct dm_target *target;
+	struct crypt_config *cc;
 	struct bio *base_bio;
 	struct work_struct work;
 
 	struct convert_context ctx;
 
-	atomic_t pending;
+	atomic_t io_pending;
 	int error;
 	sector_t sector;
 	struct dm_crypt_io *base_io;
@@ -109,9 +109,6 @@ enum flags { DM_CRYPT_SUSPENDED, DM_CRYPT_KEY_VALID };
  */
 struct crypt_cpu {
 	struct ablkcipher_request *req;
-	/* ESSIV: struct crypto_cipher *essiv_tfm */
-	void *iv_private;
-	struct crypto_ablkcipher *tfms[0];
 };
 
 /*
@@ -151,6 +148,10 @@ struct crypt_config {
 	 * per_cpu_ptr() only.
 	 */
 	struct crypt_cpu __percpu *cpu;
+
+	/* ESSIV: struct crypto_cipher *essiv_tfm */
+	void *iv_private;
+	struct crypto_ablkcipher **tfms;
 	unsigned tfms_count;
 
 	/*
@@ -193,7 +194,7 @@ static struct crypt_cpu *this_crypt_config(struct crypt_config *cc)
  */
 static struct crypto_ablkcipher *any_tfm(struct crypt_config *cc)
 {
-	return __this_cpu_ptr(cc->cpu)->tfms[0];
+	return cc->tfms[0];
 }
 
 /*
@@ -258,7 +259,7 @@ static int crypt_iv_essiv_init(struct crypt_config *cc)
 	struct hash_desc desc;
 	struct scatterlist sg;
 	struct crypto_cipher *essiv_tfm;
-	int err, cpu;
+	int err;
 
 	sg_init_one(&sg, cc->key, cc->key_size);
 	desc.tfm = essiv->hash_tfm;
@@ -268,14 +269,12 @@ static int crypt_iv_essiv_init(struct crypt_config *cc)
 	if (err)
 		return err;
 
-	for_each_possible_cpu(cpu) {
-		essiv_tfm = per_cpu_ptr(cc->cpu, cpu)->iv_private,
+	essiv_tfm = cc->iv_private;
 
-		err = crypto_cipher_setkey(essiv_tfm, essiv->salt,
-				    crypto_hash_digestsize(essiv->hash_tfm));
-		if (err)
-			return err;
-	}
+	err = crypto_cipher_setkey(essiv_tfm, essiv->salt,
+			    crypto_hash_digestsize(essiv->hash_tfm));
+	if (err)
+		return err;
 
 	return 0;
 }
@@ -286,16 +285,14 @@ 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);
 	struct crypto_cipher *essiv_tfm;
-	int cpu, r, err = 0;
+	int r, err = 0;
 
 	memset(essiv->salt, 0, salt_size);
 
-	for_each_possible_cpu(cpu) {
-		essiv_tfm = per_cpu_ptr(cc->cpu, cpu)->iv_private;
-		r = crypto_cipher_setkey(essiv_tfm, essiv->salt, salt_size);
-		if (r)
-			err = r;
-	}
+	essiv_tfm = cc->iv_private;
+	r = crypto_cipher_setkey(essiv_tfm, essiv->salt, salt_size);
+	if (r)
+		err = r;
 
 	return err;
 }
@@ -335,8 +332,6 @@ static struct crypto_cipher *setup_essiv_cpu(struct crypt_config *cc,
 
 static void crypt_iv_essiv_dtr(struct crypt_config *cc)
 {
-	int cpu;
-	struct crypt_cpu *cpu_cc;
 	struct crypto_cipher *essiv_tfm;
 	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
 
@@ -346,15 +341,12 @@ static void crypt_iv_essiv_dtr(struct crypt_config *cc)
 	kzfree(essiv->salt);
 	essiv->salt = NULL;
 
-	for_each_possible_cpu(cpu) {
-		cpu_cc = per_cpu_ptr(cc->cpu, cpu);
-		essiv_tfm = cpu_cc->iv_private;
+	essiv_tfm = cc->iv_private;
 
-		if (essiv_tfm)
-			crypto_free_cipher(essiv_tfm);
+	if (essiv_tfm)
+		crypto_free_cipher(essiv_tfm);
 
-		cpu_cc->iv_private = NULL;
-	}
+	cc->iv_private = NULL;
 }
 
 static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
@@ -363,7 +355,7 @@ static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
 	struct crypto_cipher *essiv_tfm = NULL;
 	struct crypto_hash *hash_tfm = NULL;
 	u8 *salt = NULL;
-	int err, cpu;
+	int err;
 
 	if (!opts) {
 		ti->error = "Digest algorithm missing for ESSIV mode";
@@ -388,15 +380,13 @@ static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
 	cc->iv_gen_private.essiv.salt = salt;
 	cc->iv_gen_private.essiv.hash_tfm = hash_tfm;
 
-	for_each_possible_cpu(cpu) {
-		essiv_tfm = setup_essiv_cpu(cc, ti, salt,
-					crypto_hash_digestsize(hash_tfm));
-		if (IS_ERR(essiv_tfm)) {
-			crypt_iv_essiv_dtr(cc);
-			return PTR_ERR(essiv_tfm);
-		}
-		per_cpu_ptr(cc->cpu, cpu)->iv_private = essiv_tfm;
+	essiv_tfm = setup_essiv_cpu(cc, ti, salt,
+				crypto_hash_digestsize(hash_tfm));
+	if (IS_ERR(essiv_tfm)) {
+		crypt_iv_essiv_dtr(cc);
+		return PTR_ERR(essiv_tfm);
 	}
+	cc->iv_private = essiv_tfm;
 
 	return 0;
 
@@ -410,7 +400,7 @@ bad:
 static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv,
 			      struct dm_crypt_request *dmreq)
 {
-	struct crypto_cipher *essiv_tfm = this_crypt_config(cc)->iv_private;
+	struct crypto_cipher *essiv_tfm = cc->iv_private;
 
 	memset(iv, 0, cc->iv_size);
 	*(__le64 *)iv = cpu_to_le64(dmreq->iv_sector);
@@ -664,7 +654,7 @@ static void crypt_convert_init(struct crypt_config *cc,
 	ctx->offset_out = 0;
 	ctx->idx_in = bio_in ? bio_in->bi_idx : 0;
 	ctx->idx_out = bio_out ? bio_out->bi_idx : 0;
-	ctx->sector = sector + cc->iv_offset;
+	ctx->cc_sector = sector + cc->iv_offset;
 	init_completion(&ctx->restart);
 }
 
@@ -695,12 +685,12 @@ static int crypt_convert_block(struct crypt_config *cc,
 	struct bio_vec *bv_out = bio_iovec_idx(ctx->bio_out, ctx->idx_out);
 	struct dm_crypt_request *dmreq;
 	u8 *iv;
-	int r = 0;
+	int r;
 
 	dmreq = dmreq_of_req(cc, req);
 	iv = iv_of_dmreq(cc, dmreq);
 
-	dmreq->iv_sector = ctx->sector;
+	dmreq->iv_sector = ctx->cc_sector;
 	dmreq->ctx = ctx;
 	sg_init_table(&dmreq->sg_in, 1);
 	sg_set_page(&dmreq->sg_in, bv_in->bv_page, 1 << SECTOR_SHIFT,
@@ -749,12 +739,12 @@ static void crypt_alloc_req(struct crypt_config *cc,
 			    struct convert_context *ctx)
 {
 	struct crypt_cpu *this_cc = this_crypt_config(cc);
-	unsigned key_index = ctx->sector & (cc->tfms_count - 1);
+	unsigned key_index = ctx->cc_sector & (cc->tfms_count - 1);
 
 	if (!this_cc->req)
 		this_cc->req = mempool_alloc(cc->req_pool, GFP_NOIO);
 
-	ablkcipher_request_set_tfm(this_cc->req, this_cc->tfms[key_index]);
+	ablkcipher_request_set_tfm(this_cc->req, cc->tfms[key_index]);
 	ablkcipher_request_set_callback(this_cc->req,
 	    CRYPTO_TFM_REQ_MAY_BACKLOG | CRYPTO_TFM_REQ_MAY_SLEEP,
 	    kcryptd_async_done, dmreq_of_req(cc, this_cc->req));
@@ -769,14 +759,14 @@ static int crypt_convert(struct crypt_config *cc,
 	struct crypt_cpu *this_cc = this_crypt_config(cc);
 	int r;
 
-	atomic_set(&ctx->pending, 1);
+	atomic_set(&ctx->cc_pending, 1);
 
 	while(ctx->idx_in < ctx->bio_in->bi_vcnt &&
 	      ctx->idx_out < ctx->bio_out->bi_vcnt) {
 
 		crypt_alloc_req(cc, ctx);
 
-		atomic_inc(&ctx->pending);
+		atomic_inc(&ctx->cc_pending);
 
 		r = crypt_convert_block(cc, ctx, this_cc->req);
 
@@ -788,19 +778,19 @@ static int crypt_convert(struct crypt_config *cc,
 			/* fall through*/
 		case -EINPROGRESS:
 			this_cc->req = NULL;
-			ctx->sector++;
+			ctx->cc_sector++;
 			continue;
 
 		/* sync */
 		case 0:
-			atomic_dec(&ctx->pending);
-			ctx->sector++;
+			atomic_dec(&ctx->cc_pending);
+			ctx->cc_sector++;
 			cond_resched();
 			continue;
 
 		/* error */
 		default:
-			atomic_dec(&ctx->pending);
+			atomic_dec(&ctx->cc_pending);
 			return r;
 		}
 	}
@@ -811,7 +801,7 @@ static int crypt_convert(struct crypt_config *cc,
 static void dm_crypt_bio_destructor(struct bio *bio)
 {
 	struct dm_crypt_io *io = bio->bi_private;
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	bio_free(bio, cc->bs);
 }
@@ -825,7 +815,7 @@ static void dm_crypt_bio_destructor(struct bio *bio)
 static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size,
 				      unsigned *out_of_pages)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	struct bio *clone;
 	unsigned int nr_iovecs = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	gfp_t gfp_mask = GFP_NOIO | __GFP_HIGHMEM;
@@ -884,26 +874,25 @@ static void crypt_free_buffer_pages(struct crypt_config *cc, struct bio *clone)
 	}
 }
 
-static struct dm_crypt_io *crypt_io_alloc(struct dm_target *ti,
+static struct dm_crypt_io *crypt_io_alloc(struct crypt_config *cc,
 					  struct bio *bio, sector_t sector)
 {
-	struct crypt_config *cc = ti->private;
 	struct dm_crypt_io *io;
 
 	io = mempool_alloc(cc->io_pool, GFP_NOIO);
-	io->target = ti;
+	io->cc = cc;
 	io->base_bio = bio;
 	io->sector = sector;
 	io->error = 0;
 	io->base_io = NULL;
-	atomic_set(&io->pending, 0);
+	atomic_set(&io->io_pending, 0);
 
 	return io;
 }
 
 static void crypt_inc_pending(struct dm_crypt_io *io)
 {
-	atomic_inc(&io->pending);
+	atomic_inc(&io->io_pending);
 }
 
 /*
@@ -913,12 +902,12 @@ static void crypt_inc_pending(struct dm_crypt_io *io)
  */
 static void crypt_dec_pending(struct dm_crypt_io *io)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	struct bio *base_bio = io->base_bio;
 	struct dm_crypt_io *base_io = io->base_io;
 	int error = io->error;
 
-	if (!atomic_dec_and_test(&io->pending))
+	if (!atomic_dec_and_test(&io->io_pending))
 		return;
 
 	mempool_free(io, cc->io_pool);
@@ -952,7 +941,7 @@ static void crypt_dec_pending(struct dm_crypt_io *io)
 static void crypt_endio(struct bio *clone, int error)
 {
 	struct dm_crypt_io *io = clone->bi_private;
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	unsigned rw = bio_data_dir(clone);
 
 	if (unlikely(!bio_flagged(clone, BIO_UPTODATE) && !error))
@@ -979,7 +968,7 @@ static void crypt_endio(struct bio *clone, int error)
 
 static void clone_init(struct dm_crypt_io *io, struct bio *clone)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	clone->bi_private = io;
 	clone->bi_end_io  = crypt_endio;
@@ -990,7 +979,7 @@ static void clone_init(struct dm_crypt_io *io, struct bio *clone)
 
 static int kcryptd_io_read(struct dm_crypt_io *io, gfp_t gfp)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	struct bio *base_bio = io->base_bio;
 	struct bio *clone;
 
@@ -1038,7 +1027,7 @@ static void kcryptd_io(struct work_struct *work)
 
 static void kcryptd_queue_io(struct dm_crypt_io *io)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	INIT_WORK(&io->work, kcryptd_io);
 	queue_work(cc->io_queue, &io->work);
@@ -1047,7 +1036,7 @@ static void kcryptd_queue_io(struct dm_crypt_io *io)
 static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io, int async)
 {
 	struct bio *clone = io->ctx.bio_out;
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	if (unlikely(io->error < 0)) {
 		crypt_free_buffer_pages(cc, clone);
@@ -1069,7 +1058,7 @@ static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io, int async)
 
 static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	struct bio *clone;
 	struct dm_crypt_io *new_io;
 	int crypt_finished;
@@ -1107,7 +1096,7 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 		if (r < 0)
 			io->error = -EIO;
 
-		crypt_finished = atomic_dec_and_test(&io->ctx.pending);
+		crypt_finished = atomic_dec_and_test(&io->ctx.cc_pending);
 
 		/* Encryption was already finished, submit io now */
 		if (crypt_finished) {
@@ -1135,7 +1124,7 @@ static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
 		 * between fragments, so switch to a new dm_crypt_io structure.
 		 */
 		if (unlikely(!crypt_finished && remaining)) {
-			new_io = crypt_io_alloc(io->target, io->base_bio,
+			new_io = crypt_io_alloc(io->cc, io->base_bio,
 						sector);
 			crypt_inc_pending(new_io);
 			crypt_convert_init(cc, &new_io->ctx, NULL,
@@ -1169,7 +1158,7 @@ static void kcryptd_crypt_read_done(struct dm_crypt_io *io)
 
 static void kcryptd_crypt_read_convert(struct dm_crypt_io *io)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 	int r = 0;
 
 	crypt_inc_pending(io);
@@ -1181,7 +1170,7 @@ static void kcryptd_crypt_read_convert(struct dm_crypt_io *io)
 	if (r < 0)
 		io->error = -EIO;
 
-	if (atomic_dec_and_test(&io->ctx.pending))
+	if (atomic_dec_and_test(&io->ctx.cc_pending))
 		kcryptd_crypt_read_done(io);
 
 	crypt_dec_pending(io);
@@ -1193,7 +1182,7 @@ static void kcryptd_async_done(struct crypto_async_request *async_req,
 	struct dm_crypt_request *dmreq = async_req->data;
 	struct convert_context *ctx = dmreq->ctx;
 	struct dm_crypt_io *io = container_of(ctx, struct dm_crypt_io, ctx);
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	if (error == -EINPROGRESS) {
 		complete(&ctx->restart);
@@ -1208,7 +1197,7 @@ static void kcryptd_async_done(struct crypto_async_request *async_req,
 
 	mempool_free(req_of_dmreq(cc, dmreq), cc->req_pool);
 
-	if (!atomic_dec_and_test(&ctx->pending))
+	if (!atomic_dec_and_test(&ctx->cc_pending))
 		return;
 
 	if (bio_data_dir(io->base_bio) == READ)
@@ -1229,7 +1218,7 @@ static void kcryptd_crypt(struct work_struct *work)
 
 static void kcryptd_queue_crypt(struct dm_crypt_io *io)
 {
-	struct crypt_config *cc = io->target->private;
+	struct crypt_config *cc = io->cc;
 
 	INIT_WORK(&io->work, kcryptd_crypt);
 	queue_work(cc->crypt_queue, &io->work);
@@ -1241,7 +1230,6 @@ static void kcryptd_queue_crypt(struct dm_crypt_io *io)
 static int crypt_decode_key(u8 *key, char *hex, unsigned int size)
 {
 	char buffer[3];
-	char *endp;
 	unsigned int i;
 
 	buffer[2] = '\0';
@@ -1250,9 +1238,7 @@ static int crypt_decode_key(u8 *key, char *hex, unsigned int size)
 		buffer[0] = *hex++;
 		buffer[1] = *hex++;
 
-		key[i] = (u8)simple_strtoul(buffer, &endp, 16);
-
-		if (endp != &buffer[2])
+		if (kstrtou8(buffer, 16, &key[i]))
 			return -EINVAL;
 	}
 
@@ -1276,29 +1262,38 @@ static void crypt_encode_key(char *hex, u8 *key, unsigned int size)
 	}
 }
 
-static void crypt_free_tfms(struct crypt_config *cc, int cpu)
+static void crypt_free_tfms(struct crypt_config *cc)
 {
-	struct crypt_cpu *cpu_cc = per_cpu_ptr(cc->cpu, cpu);
 	unsigned i;
 
+	if (!cc->tfms)
+		return;
+
 	for (i = 0; i < cc->tfms_count; i++)
-		if (cpu_cc->tfms[i] && !IS_ERR(cpu_cc->tfms[i])) {
-			crypto_free_ablkcipher(cpu_cc->tfms[i]);
-			cpu_cc->tfms[i] = NULL;
+		if (cc->tfms[i] && !IS_ERR(cc->tfms[i])) {
+			crypto_free_ablkcipher(cc->tfms[i]);
+			cc->tfms[i] = NULL;
 		}
+
+	kfree(cc->tfms);
+	cc->tfms = NULL;
 }
 
-static int crypt_alloc_tfms(struct crypt_config *cc, int cpu, char *ciphermode)
+static int crypt_alloc_tfms(struct crypt_config *cc, char *ciphermode)
 {
-	struct crypt_cpu *cpu_cc = per_cpu_ptr(cc->cpu, cpu);
 	unsigned i;
 	int err;
 
+	cc->tfms = kmalloc(cc->tfms_count * sizeof(struct crypto_ablkcipher *),
+			   GFP_KERNEL);
+	if (!cc->tfms)
+		return -ENOMEM;
+
 	for (i = 0; i < cc->tfms_count; i++) {
-		cpu_cc->tfms[i] = crypto_alloc_ablkcipher(ciphermode, 0, 0);
-		if (IS_ERR(cpu_cc->tfms[i])) {
-			err = PTR_ERR(cpu_cc->tfms[i]);
-			crypt_free_tfms(cc, cpu);
+		cc->tfms[i] = crypto_alloc_ablkcipher(ciphermode, 0, 0);
+		if (IS_ERR(cc->tfms[i])) {
+			err = PTR_ERR(cc->tfms[i]);
+			crypt_free_tfms(cc);
 			return err;
 		}
 	}
@@ -1309,15 +1304,14 @@ static int crypt_alloc_tfms(struct crypt_config *cc, int cpu, char *ciphermode)
 static int crypt_setkey_allcpus(struct crypt_config *cc)
 {
 	unsigned subkey_size = cc->key_size >> ilog2(cc->tfms_count);
-	int cpu, err = 0, i, r;
-
-	for_each_possible_cpu(cpu) {
-		for (i = 0; i < cc->tfms_count; i++) {
-			r = crypto_ablkcipher_setkey(per_cpu_ptr(cc->cpu, cpu)->tfms[i],
-						     cc->key + (i * subkey_size), subkey_size);
-			if (r)
-				err = r;
-		}
+	int err = 0, i, r;
+
+	for (i = 0; i < cc->tfms_count; i++) {
+		r = crypto_ablkcipher_setkey(cc->tfms[i],
+					     cc->key + (i * subkey_size),
+					     subkey_size);
+		if (r)
+			err = r;
 	}
 
 	return err;
@@ -1379,9 +1373,10 @@ static void crypt_dtr(struct dm_target *ti)
 			cpu_cc = per_cpu_ptr(cc->cpu, cpu);
 			if (cpu_cc->req)
 				mempool_free(cpu_cc->req, cc->req_pool);
-			crypt_free_tfms(cc, cpu);
 		}
 
+	crypt_free_tfms(cc);
+
 	if (cc->bs)
 		bioset_free(cc->bs);
 
@@ -1414,7 +1409,7 @@ static int crypt_ctr_cipher(struct dm_target *ti,
 	struct crypt_config *cc = ti->private;
 	char *tmp, *cipher, *chainmode, *ivmode, *ivopts, *keycount;
 	char *cipher_api = NULL;
-	int cpu, ret = -EINVAL;
+	int ret = -EINVAL;
 	char dummy;
 
 	/* Convert to crypto api definition? */
@@ -1455,8 +1450,7 @@ static int crypt_ctr_cipher(struct dm_target *ti,
 	if (tmp)
 		DMWARN("Ignoring unexpected additional cipher options");
 
-	cc->cpu = __alloc_percpu(sizeof(*(cc->cpu)) +
-				 cc->tfms_count * sizeof(*(cc->cpu->tfms)),
+	cc->cpu = __alloc_percpu(sizeof(*(cc->cpu)),
 				 __alignof__(struct crypt_cpu));
 	if (!cc->cpu) {
 		ti->error = "Cannot allocate per cpu state";
@@ -1489,12 +1483,10 @@ static int crypt_ctr_cipher(struct dm_target *ti,
 	}
 
 	/* Allocate cipher */
-	for_each_possible_cpu(cpu) {
-		ret = crypt_alloc_tfms(cc, cpu, cipher_api);
-		if (ret < 0) {
-			ti->error = "Error allocating crypto tfm";
-			goto bad;
-		}
+	ret = crypt_alloc_tfms(cc, cipher_api);
+	if (ret < 0) {
+		ti->error = "Error allocating crypto tfm";
+		goto bad;
 	}
 
 	/* Initialize and set key */
@@ -1702,7 +1694,7 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	}
 
 	ti->num_flush_requests = 1;
-	ti->discard_zeroes_data_unsupported = 1;
+	ti->discard_zeroes_data_unsupported = true;
 
 	return 0;
 
@@ -1715,7 +1707,7 @@ static int crypt_map(struct dm_target *ti, struct bio *bio,
 		     union map_info *map_context)
 {
 	struct dm_crypt_io *io;
-	struct crypt_config *cc;
+	struct crypt_config *cc = ti->private;
 
 	/*
 	 * If bio is REQ_FLUSH or REQ_DISCARD, just bypass crypt queues.
@@ -1723,14 +1715,13 @@ static int crypt_map(struct dm_target *ti, struct bio *bio,
 	 * - for REQ_DISCARD caller must use flush if IO ordering matters
 	 */
 	if (unlikely(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD))) {
-		cc = ti->private;
 		bio->bi_bdev = cc->dev->bdev;
 		if (bio_sectors(bio))
 			bio->bi_sector = cc->start + dm_target_offset(ti, bio->bi_sector);
 		return DM_MAPIO_REMAPPED;
 	}
 
-	io = crypt_io_alloc(ti, bio, dm_target_offset(ti, bio->bi_sector));
+	io = crypt_io_alloc(cc, bio, dm_target_offset(ti, bio->bi_sector));
 
 	if (bio_data_dir(io->base_bio) == READ) {
 		if (kcryptd_io_read(io, GFP_NOWAIT))
@@ -1742,7 +1733,7 @@ static int crypt_map(struct dm_target *ti, struct bio *bio,
 }
 
 static int crypt_status(struct dm_target *ti, status_type_t type,
-			char *result, unsigned int maxlen)
+			unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct crypt_config *cc = ti->private;
 	unsigned int sz = 0;
diff --git a/drivers/md/dm-delay.c b/drivers/md/dm-delay.c
index 2dc22dddb2a..f53846f9ab5 100644
--- a/drivers/md/dm-delay.c
+++ b/drivers/md/dm-delay.c
@@ -295,7 +295,7 @@ static int delay_map(struct dm_target *ti, struct bio *bio,
 }
 
 static int delay_status(struct dm_target *ti, status_type_t type,
-			char *result, unsigned maxlen)
+			unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct delay_c *dc = ti->private;
 	int sz = 0;
diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index aa70f7d43a1..ebaa4f803ee 100644
--- a/drivers/md/dm-exception-store.c
+++ b/drivers/md/dm-exception-store.c
@@ -142,24 +142,19 @@ EXPORT_SYMBOL(dm_exception_store_type_unregister);
 static int set_chunk_size(struct dm_exception_store *store,
 			  const char *chunk_size_arg, char **error)
 {
-	unsigned long chunk_size_ulong;
-	char *value;
+	unsigned chunk_size;
 
-	chunk_size_ulong = simple_strtoul(chunk_size_arg, &value, 10);
-	if (*chunk_size_arg == '\0' || *value != '\0' ||
-	    chunk_size_ulong > UINT_MAX) {
+	if (kstrtouint(chunk_size_arg, 10, &chunk_size)) {
 		*error = "Invalid chunk size";
 		return -EINVAL;
 	}
 
-	if (!chunk_size_ulong) {
+	if (!chunk_size) {
 		store->chunk_size = store->chunk_mask = store->chunk_shift = 0;
 		return 0;
 	}
 
-	return dm_exception_store_set_chunk_size(store,
-						 (unsigned) chunk_size_ulong,
-						 error);
+	return dm_exception_store_set_chunk_size(store, chunk_size, error);
 }
 
 int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
diff --git a/drivers/md/dm-flakey.c b/drivers/md/dm-flakey.c
index ac49c01f1a4..cc15543a6ad 100644
--- a/drivers/md/dm-flakey.c
+++ b/drivers/md/dm-flakey.c
@@ -333,7 +333,7 @@ static int flakey_end_io(struct dm_target *ti, struct bio *bio,
 }
 
 static int flakey_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned int maxlen)
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	unsigned sz = 0;
 	struct flakey_c *fc = ti->private;
diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
index a1a3e6df17b..afd95986d09 100644
--- a/drivers/md/dm-ioctl.c
+++ b/drivers/md/dm-ioctl.c
@@ -1054,6 +1054,7 @@ static void retrieve_status(struct dm_table *table,
 	char *outbuf, *outptr;
 	status_type_t type;
 	size_t remaining, len, used = 0;
+	unsigned status_flags = 0;
 
 	outptr = outbuf = get_result_buffer(param, param_size, &len);
 
@@ -1090,7 +1091,9 @@ static void retrieve_status(struct dm_table *table,
 
 		/* Get the status/table string from the target driver */
 		if (ti->type->status) {
-			if (ti->type->status(ti, type, outptr, remaining)) {
+			if (param->flags & DM_NOFLUSH_FLAG)
+				status_flags |= DM_STATUS_NOFLUSH_FLAG;
+			if (ti->type->status(ti, type, status_flags, outptr, remaining)) {
 				param->flags |= DM_BUFFER_FULL_FLAG;
 				break;
 			}
diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c
index 3639eeab604..1bf19a93eef 100644
--- a/drivers/md/dm-linear.c
+++ b/drivers/md/dm-linear.c
@@ -96,7 +96,7 @@ static int linear_map(struct dm_target *ti, struct bio *bio,
 }
 
 static int linear_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned int maxlen)
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct linear_c *lc = (struct linear_c *) ti->private;
 
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index 65ebaebf502..627d19186d5 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -571,16 +571,6 @@ static void disk_dtr(struct dm_dirty_log *log)
 	destroy_log_context(lc);
 }
 
-static int count_bits32(uint32_t *addr, unsigned size)
-{
-	int count = 0, i;
-
-	for (i = 0; i < size; i++) {
-		count += hweight32(*(addr+i));
-	}
-	return count;
-}
-
 static void fail_log_device(struct log_c *lc)
 {
 	if (lc->log_dev_failed)
@@ -629,7 +619,8 @@ static int disk_resume(struct dm_dirty_log *log)
 
 	/* copy clean across to sync */
 	memcpy(lc->sync_bits, lc->clean_bits, size);
-	lc->sync_count = count_bits32(lc->clean_bits, lc->bitset_uint32_count);
+	lc->sync_count = memweight(lc->clean_bits,
+				lc->bitset_uint32_count * sizeof(uint32_t));
 	lc->sync_search = 0;
 
 	/* set the correct number of regions in the header */
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 638dae048b4..d8abb90a6c2 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -85,6 +85,7 @@ struct multipath {
 	unsigned queue_io:1;		/* Must we queue all I/O? */
 	unsigned queue_if_no_path:1;	/* Queue I/O if last path fails? */
 	unsigned saved_queue_if_no_path:1; /* Saved state during suspension */
+	unsigned retain_attached_hw_handler:1; /* If there's already a hw_handler present, don't change it. */
 
 	unsigned pg_init_retries;	/* Number of times to retry pg_init */
 	unsigned pg_init_count;		/* Number of times pg_init called */
@@ -568,6 +569,8 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
 	int r;
 	struct pgpath *p;
 	struct multipath *m = ti->private;
+	struct request_queue *q = NULL;
+	const char *attached_handler_name;
 
 	/* we need at least a path arg */
 	if (as->argc < 1) {
@@ -586,13 +589,37 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
 		goto bad;
 	}
 
-	if (m->hw_handler_name) {
-		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
+	if (m->retain_attached_hw_handler || m->hw_handler_name)
+		q = bdev_get_queue(p->path.dev->bdev);
+
+	if (m->retain_attached_hw_handler) {
+		attached_handler_name = scsi_dh_attached_handler_name(q, GFP_KERNEL);
+		if (attached_handler_name) {
+			/*
+			 * Reset hw_handler_name to match the attached handler
+			 * and clear any hw_handler_params associated with the
+			 * ignored handler.
+			 *
+			 * NB. This modifies the table line to show the actual
+			 * handler instead of the original table passed in.
+			 */
+			kfree(m->hw_handler_name);
+			m->hw_handler_name = attached_handler_name;
+
+			kfree(m->hw_handler_params);
+			m->hw_handler_params = NULL;
+		}
+	}
 
+	if (m->hw_handler_name) {
+		/*
+		 * Increments scsi_dh reference, even when using an
+		 * already-attached handler.
+		 */
 		r = scsi_dh_attach(q, m->hw_handler_name);
 		if (r == -EBUSY) {
 			/*
-			 * Already attached to different hw_handler,
+			 * Already attached to different hw_handler:
 			 * try to reattach with correct one.
 			 */
 			scsi_dh_detach(q);
@@ -760,7 +787,7 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m)
 	const char *arg_name;
 
 	static struct dm_arg _args[] = {
-		{0, 5, "invalid number of feature args"},
+		{0, 6, "invalid number of feature args"},
 		{1, 50, "pg_init_retries must be between 1 and 50"},
 		{0, 60000, "pg_init_delay_msecs must be between 0 and 60000"},
 	};
@@ -781,6 +808,11 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m)
 			continue;
 		}
 
+		if (!strcasecmp(arg_name, "retain_attached_hw_handler")) {
+			m->retain_attached_hw_handler = 1;
+			continue;
+		}
+
 		if (!strcasecmp(arg_name, "pg_init_retries") &&
 		    (argc >= 1)) {
 			r = dm_read_arg(_args + 1, as, &m->pg_init_retries, &ti->error);
@@ -1346,7 +1378,7 @@ static void multipath_resume(struct dm_target *ti)
  *      num_paths num_selector_args [path_dev [selector_args]* ]+ ]+
  */
 static int multipath_status(struct dm_target *ti, status_type_t type,
-			    char *result, unsigned int maxlen)
+			    unsigned status_flags, char *result, unsigned maxlen)
 {
 	int sz = 0;
 	unsigned long flags;
@@ -1364,13 +1396,16 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 	else {
 		DMEMIT("%u ", m->queue_if_no_path +
 			      (m->pg_init_retries > 0) * 2 +
-			      (m->pg_init_delay_msecs != DM_PG_INIT_DELAY_DEFAULT) * 2);
+			      (m->pg_init_delay_msecs != DM_PG_INIT_DELAY_DEFAULT) * 2 +
+			      m->retain_attached_hw_handler);
 		if (m->queue_if_no_path)
 			DMEMIT("queue_if_no_path ");
 		if (m->pg_init_retries)
 			DMEMIT("pg_init_retries %u ", m->pg_init_retries);
 		if (m->pg_init_delay_msecs != DM_PG_INIT_DELAY_DEFAULT)
 			DMEMIT("pg_init_delay_msecs %u ", m->pg_init_delay_msecs);
+		if (m->retain_attached_hw_handler)
+			DMEMIT("retain_attached_hw_handler ");
 	}
 
 	if (!m->hw_handler_name || type == STATUSTYPE_INFO)
@@ -1656,7 +1691,7 @@ out:
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 4, 0},
+	.version = {1, 5, 0},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,
diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
index 017c34d78d6..982e3e390c4 100644
--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -11,6 +11,7 @@
 #include "md.h"
 #include "raid1.h"
 #include "raid5.h"
+#include "raid10.h"
 #include "bitmap.h"
 
 #include <linux/device-mapper.h>
@@ -52,7 +53,10 @@ struct raid_dev {
 #define DMPF_MAX_RECOVERY_RATE 0x20
 #define DMPF_MAX_WRITE_BEHIND  0x40
 #define DMPF_STRIPE_CACHE      0x80
-#define DMPF_REGION_SIZE       0X100
+#define DMPF_REGION_SIZE       0x100
+#define DMPF_RAID10_COPIES     0x200
+#define DMPF_RAID10_FORMAT     0x400
+
 struct raid_set {
 	struct dm_target *ti;
 
@@ -76,6 +80,7 @@ static struct raid_type {
 	const unsigned algorithm;	/* RAID algorithm. */
 } raid_types[] = {
 	{"raid1",    "RAID1 (mirroring)",               0, 2, 1, 0 /* NONE */},
+	{"raid10",   "RAID10 (striped mirrors)",        0, 2, 10, UINT_MAX /* Varies */},
 	{"raid4",    "RAID4 (dedicated parity disk)",	1, 2, 5, ALGORITHM_PARITY_0},
 	{"raid5_la", "RAID5 (left asymmetric)",		1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC},
 	{"raid5_ra", "RAID5 (right asymmetric)",	1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC},
@@ -86,6 +91,17 @@ static struct raid_type {
 	{"raid6_nc", "RAID6 (N continue)",		2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}
 };
 
+static unsigned raid10_md_layout_to_copies(int layout)
+{
+	return layout & 0xFF;
+}
+
+static int raid10_format_to_md_layout(char *format, unsigned copies)
+{
+	/* 1 "far" copy, and 'copies' "near" copies */
+	return (1 << 8) | (copies & 0xFF);
+}
+
 static struct raid_type *get_raid_type(char *name)
 {
 	int i;
@@ -101,20 +117,12 @@ static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *ra
 {
 	unsigned i;
 	struct raid_set *rs;
-	sector_t sectors_per_dev;
 
 	if (raid_devs <= raid_type->parity_devs) {
 		ti->error = "Insufficient number of devices";
 		return ERR_PTR(-EINVAL);
 	}
 
-	sectors_per_dev = ti->len;
-	if ((raid_type->level > 1) &&
-	    sector_div(sectors_per_dev, (raid_devs - raid_type->parity_devs))) {
-		ti->error = "Target length not divisible by number of data devices";
-		return ERR_PTR(-EINVAL);
-	}
-
 	rs = kzalloc(sizeof(*rs) + raid_devs * sizeof(rs->dev[0]), GFP_KERNEL);
 	if (!rs) {
 		ti->error = "Cannot allocate raid context";
@@ -128,7 +136,6 @@ static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *ra
 	rs->md.raid_disks = raid_devs;
 	rs->md.level = raid_type->level;
 	rs->md.new_level = rs->md.level;
-	rs->md.dev_sectors = sectors_per_dev;
 	rs->md.layout = raid_type->algorithm;
 	rs->md.new_layout = rs->md.layout;
 	rs->md.delta_disks = 0;
@@ -143,6 +150,7 @@ static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *ra
 	 *  rs->md.external
 	 *  rs->md.chunk_sectors
 	 *  rs->md.new_chunk_sectors
+	 *  rs->md.dev_sectors
 	 */
 
 	return rs;
@@ -347,12 +355,20 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size)
  *    [max_write_behind <sectors>]	See '-write-behind=' (man mdadm)
  *    [stripe_cache <sectors>]		Stripe cache size for higher RAIDs
  *    [region_size <sectors>]           Defines granularity of bitmap
+ *
+ * RAID10-only options:
+ *    [raid10_copies <# copies>]        Number of copies.  (Default: 2)
+ *    [raid10_format <near>]            Layout algorithm.  (Default: near)
  */
 static int parse_raid_params(struct raid_set *rs, char **argv,
 			     unsigned num_raid_params)
 {
+	char *raid10_format = "near";
+	unsigned raid10_copies = 2;
 	unsigned i, rebuild_cnt = 0;
 	unsigned long value, region_size = 0;
+	sector_t sectors_per_dev = rs->ti->len;
+	sector_t max_io_len;
 	char *key;
 
 	/*
@@ -422,20 +438,53 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 		}
 
 		key = argv[i++];
+
+		/* Parameters that take a string value are checked here. */
+		if (!strcasecmp(key, "raid10_format")) {
+			if (rs->raid_type->level != 10) {
+				rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type";
+				return -EINVAL;
+			}
+			if (strcmp("near", argv[i])) {
+				rs->ti->error = "Invalid 'raid10_format' value given";
+				return -EINVAL;
+			}
+			raid10_format = argv[i];
+			rs->print_flags |= DMPF_RAID10_FORMAT;
+			continue;
+		}
+
 		if (strict_strtoul(argv[i], 10, &value) < 0) {
 			rs->ti->error = "Bad numerical argument given in raid params";
 			return -EINVAL;
 		}
 
+		/* Parameters that take a numeric value are checked here */
 		if (!strcasecmp(key, "rebuild")) {
 			rebuild_cnt++;
-			if (((rs->raid_type->level != 1) &&
-			     (rebuild_cnt > rs->raid_type->parity_devs)) ||
-			    ((rs->raid_type->level == 1) &&
-			     (rebuild_cnt > (rs->md.raid_disks - 1)))) {
-				rs->ti->error = "Too many rebuild devices specified for given RAID type";
+
+			switch (rs->raid_type->level) {
+			case 1:
+				if (rebuild_cnt >= rs->md.raid_disks) {
+					rs->ti->error = "Too many rebuild devices specified";
+					return -EINVAL;
+				}
+				break;
+			case 4:
+			case 5:
+			case 6:
+				if (rebuild_cnt > rs->raid_type->parity_devs) {
+					rs->ti->error = "Too many rebuild devices specified for given RAID type";
+					return -EINVAL;
+				}
+				break;
+			case 10:
+			default:
+				DMERR("The rebuild parameter is not supported for %s", rs->raid_type->name);
+				rs->ti->error = "Rebuild not supported for this RAID type";
 				return -EINVAL;
 			}
+
 			if (value > rs->md.raid_disks) {
 				rs->ti->error = "Invalid rebuild index given";
 				return -EINVAL;
@@ -486,7 +535,8 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 			 */
 			value /= 2;
 
-			if (rs->raid_type->level < 5) {
+			if ((rs->raid_type->level != 5) &&
+			    (rs->raid_type->level != 6)) {
 				rs->ti->error = "Inappropriate argument: stripe_cache";
 				return -EINVAL;
 			}
@@ -511,6 +561,14 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 		} else if (!strcasecmp(key, "region_size")) {
 			rs->print_flags |= DMPF_REGION_SIZE;
 			region_size = value;
+		} else if (!strcasecmp(key, "raid10_copies") &&
+			   (rs->raid_type->level == 10)) {
+			if ((value < 2) || (value > 0xFF)) {
+				rs->ti->error = "Bad value for 'raid10_copies'";
+				return -EINVAL;
+			}
+			rs->print_flags |= DMPF_RAID10_COPIES;
+			raid10_copies = value;
 		} else {
 			DMERR("Unable to parse RAID parameter: %s", key);
 			rs->ti->error = "Unable to parse RAID parameters";
@@ -522,14 +580,33 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 		return -EINVAL;
 
 	if (rs->md.chunk_sectors)
-		rs->ti->split_io = rs->md.chunk_sectors;
+		max_io_len = rs->md.chunk_sectors;
 	else
-		rs->ti->split_io = region_size;
+		max_io_len = region_size;
 
-	if (rs->md.chunk_sectors)
-		rs->ti->split_io = rs->md.chunk_sectors;
-	else
-		rs->ti->split_io = region_size;
+	if (dm_set_target_max_io_len(rs->ti, max_io_len))
+		return -EINVAL;
+
+	if (rs->raid_type->level == 10) {
+		if (raid10_copies > rs->md.raid_disks) {
+			rs->ti->error = "Not enough devices to satisfy specification";
+			return -EINVAL;
+		}
+
+		/* (Len * #mirrors) / #devices */
+		sectors_per_dev = rs->ti->len * raid10_copies;
+		sector_div(sectors_per_dev, rs->md.raid_disks);
+
+		rs->md.layout = raid10_format_to_md_layout(raid10_format,
+							   raid10_copies);
+		rs->md.new_layout = rs->md.layout;
+	} else if ((rs->raid_type->level > 1) &&
+		   sector_div(sectors_per_dev,
+			      (rs->md.raid_disks - rs->raid_type->parity_devs))) {
+		rs->ti->error = "Target length not divisible by number of data devices";
+		return -EINVAL;
+	}
+	rs->md.dev_sectors = sectors_per_dev;
 
 	/* Assume there are no metadata devices until the drives are parsed */
 	rs->md.persistent = 0;
@@ -552,6 +629,9 @@ static int raid_is_congested(struct dm_target_callbacks *cb, int bits)
 	if (rs->raid_type->level == 1)
 		return md_raid1_congested(&rs->md, bits);
 
+	if (rs->raid_type->level == 10)
+		return md_raid10_congested(&rs->md, bits);
+
 	return md_raid5_congested(&rs->md, bits);
 }
 
@@ -870,6 +950,9 @@ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs)
 	case 6:
 		redundancy = rs->raid_type->parity_devs;
 		break;
+	case 10:
+		redundancy = raid10_md_layout_to_copies(mddev->layout) - 1;
+		break;
 	default:
 		ti->error = "Unknown RAID type";
 		return -EINVAL;
@@ -1035,12 +1118,19 @@ static int raid_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		goto bad;
 	}
 
+	if (ti->len != rs->md.array_sectors) {
+		ti->error = "Array size does not match requested target length";
+		ret = -EINVAL;
+		goto size_mismatch;
+	}
 	rs->callbacks.congested_fn = raid_is_congested;
 	dm_table_add_target_callbacks(ti->table, &rs->callbacks);
 
 	mddev_suspend(&rs->md);
 	return 0;
 
+size_mismatch:
+	md_stop(&rs->md);
 bad:
 	context_free(rs);
 
@@ -1067,7 +1157,7 @@ static int raid_map(struct dm_target *ti, struct bio *bio, union map_info *map_c
 }
 
 static int raid_status(struct dm_target *ti, status_type_t type,
-		       char *result, unsigned maxlen)
+		       unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct raid_set *rs = ti->private;
 	unsigned raid_param_cnt = 1; /* at least 1 for chunksize */
@@ -1189,6 +1279,13 @@ static int raid_status(struct dm_target *ti, status_type_t type,
 			DMEMIT(" region_size %lu",
 			       rs->md.bitmap_info.chunksize >> 9);
 
+		if (rs->print_flags & DMPF_RAID10_COPIES)
+			DMEMIT(" raid10_copies %u",
+			       raid10_md_layout_to_copies(rs->md.layout));
+
+		if (rs->print_flags & DMPF_RAID10_FORMAT)
+			DMEMIT(" raid10_format near");
+
 		DMEMIT(" %d", rs->md.raid_disks);
 		for (i = 0; i < rs->md.raid_disks; i++) {
 			if (rs->dev[i].meta_dev)
@@ -1263,7 +1360,7 @@ static void raid_resume(struct dm_target *ti)
 
 static struct target_type raid_target = {
 	.name = "raid",
-	.version = {1, 2, 0},
+	.version = {1, 3, 0},
 	.module = THIS_MODULE,
 	.ctr = raid_ctr,
 	.dtr = raid_dtr,
@@ -1290,6 +1387,8 @@ module_init(dm_raid_init);
 module_exit(dm_raid_exit);
 
 MODULE_DESCRIPTION(DM_NAME " raid4/5/6 target");
+MODULE_ALIAS("dm-raid1");
+MODULE_ALIAS("dm-raid10");
 MODULE_ALIAS("dm-raid4");
 MODULE_ALIAS("dm-raid5");
 MODULE_ALIAS("dm-raid6");
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index d039de8322f..bc5ddba8045 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -1081,9 +1081,14 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	}
 
 	ti->private = ms;
-	ti->split_io = dm_rh_get_region_size(ms->rh);
+
+	r = dm_set_target_max_io_len(ti, dm_rh_get_region_size(ms->rh));
+	if (r)
+		goto err_free_context;
+
 	ti->num_flush_requests = 1;
 	ti->num_discard_requests = 1;
+	ti->discard_zeroes_data_unsupported = true;
 
 	ms->kmirrord_wq = alloc_workqueue("kmirrord",
 					  WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0);
@@ -1214,7 +1219,7 @@ static int mirror_end_io(struct dm_target *ti, struct bio *bio,
 	 * We need to dec pending if this was a write.
 	 */
 	if (rw == WRITE) {
-		if (!(bio->bi_rw & REQ_FLUSH))
+		if (!(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD)))
 			dm_rh_dec(ms->rh, map_context->ll);
 		return error;
 	}
@@ -1362,7 +1367,7 @@ static char device_status_char(struct mirror *m)
 
 
 static int mirror_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned int maxlen)
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	unsigned int m, sz = 0;
 	struct mirror_set *ms = (struct mirror_set *) ti->private;
diff --git a/drivers/md/dm-region-hash.c b/drivers/md/dm-region-hash.c
index 7771ed21218..69732e03eb3 100644
--- a/drivers/md/dm-region-hash.c
+++ b/drivers/md/dm-region-hash.c
@@ -404,6 +404,9 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh, struct bio *bio)
 		return;
 	}
 
+	if (bio->bi_rw & REQ_DISCARD)
+		return;
+
 	/* We must inform the log that the sync count has changed. */
 	log->type->set_region_sync(log, region, 0);
 
@@ -524,7 +527,7 @@ void dm_rh_inc_pending(struct dm_region_hash *rh, struct bio_list *bios)
 	struct bio *bio;
 
 	for (bio = bios->head; bio; bio = bio->bi_next) {
-		if (bio->bi_rw & REQ_FLUSH)
+		if (bio->bi_rw & (REQ_FLUSH | REQ_DISCARD))
 			continue;
 		rh_inc(rh, dm_rh_bio_to_region(rh, bio));
 	}
diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c
index 6f758870fc1..a143921feaf 100644
--- a/drivers/md/dm-snap.c
+++ b/drivers/md/dm-snap.c
@@ -691,7 +691,7 @@ static int dm_add_exception(void *context, chunk_t old, chunk_t new)
  * 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)
+static uint32_t __minimum_chunk_size(struct origin *o)
 {
 	struct dm_snapshot *snap;
 	unsigned chunk_size = 0;
@@ -701,7 +701,7 @@ static sector_t __minimum_chunk_size(struct origin *o)
 			chunk_size = min_not_zero(chunk_size,
 						  snap->store->chunk_size);
 
-	return chunk_size;
+	return (uint32_t) chunk_size;
 }
 
 /*
@@ -1172,7 +1172,10 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		ti->error = "Chunk size not set";
 		goto bad_read_metadata;
 	}
-	ti->split_io = s->store->chunk_size;
+
+	r = dm_set_target_max_io_len(ti, s->store->chunk_size);
+	if (r)
+		goto bad_read_metadata;
 
 	return 0;
 
@@ -1239,7 +1242,7 @@ static void __handover_exceptions(struct dm_snapshot *snap_src,
 	snap_dest->store->snap = snap_dest;
 	snap_src->store->snap = snap_src;
 
-	snap_dest->ti->split_io = snap_dest->store->chunk_size;
+	snap_dest->ti->max_io_len = snap_dest->store->chunk_size;
 	snap_dest->valid = snap_src->valid;
 
 	/*
@@ -1817,9 +1820,9 @@ static void snapshot_resume(struct dm_target *ti)
 	up_write(&s->lock);
 }
 
-static sector_t get_origin_minimum_chunksize(struct block_device *bdev)
+static uint32_t get_origin_minimum_chunksize(struct block_device *bdev)
 {
-	sector_t min_chunksize;
+	uint32_t min_chunksize;
 
 	down_read(&_origins_lock);
 	min_chunksize = __minimum_chunk_size(__lookup_origin(bdev));
@@ -1838,15 +1841,15 @@ static void snapshot_merge_resume(struct dm_target *ti)
 	snapshot_resume(ti);
 
 	/*
-	 * snapshot-merge acts as an origin, so set ti->split_io
+	 * snapshot-merge acts as an origin, so set ti->max_io_len
 	 */
-	ti->split_io = get_origin_minimum_chunksize(s->origin->bdev);
+	ti->max_io_len = get_origin_minimum_chunksize(s->origin->bdev);
 
 	start_merge(s);
 }
 
 static int snapshot_status(struct dm_target *ti, status_type_t type,
-			   char *result, unsigned int maxlen)
+			   unsigned status_flags, char *result, unsigned maxlen)
 {
 	unsigned sz = 0;
 	struct dm_snapshot *snap = ti->private;
@@ -2073,12 +2076,12 @@ static int origin_write_extent(struct dm_snapshot *merging_snap,
 	struct origin *o;
 
 	/*
-	 * The origin's __minimum_chunk_size() got stored in split_io
+	 * The origin's __minimum_chunk_size() got stored in max_io_len
 	 * by snapshot_merge_resume().
 	 */
 	down_read(&_origins_lock);
 	o = __lookup_origin(merging_snap->origin->bdev);
-	for (n = 0; n < size; n += merging_snap->ti->split_io)
+	for (n = 0; n < size; n += merging_snap->ti->max_io_len)
 		if (__origin_write(&o->snapshots, sector + n, NULL) ==
 		    DM_MAPIO_SUBMITTED)
 			must_wait = 1;
@@ -2138,18 +2141,18 @@ static int origin_map(struct dm_target *ti, struct bio *bio,
 }
 
 /*
- * Set the target "split_io" field to the minimum of all the snapshots'
+ * Set the target "max_io_len" field to the minimum of all the snapshots'
  * chunk sizes.
  */
 static void origin_resume(struct dm_target *ti)
 {
 	struct dm_dev *dev = ti->private;
 
-	ti->split_io = get_origin_minimum_chunksize(dev->bdev);
+	ti->max_io_len = get_origin_minimum_chunksize(dev->bdev);
 }
 
-static int origin_status(struct dm_target *ti, status_type_t type, char *result,
-			 unsigned int maxlen)
+static int origin_status(struct dm_target *ti, status_type_t type,
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct dm_dev *dev = ti->private;
 
@@ -2176,7 +2179,6 @@ static int origin_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
 		return max_size;
 
 	bvm->bi_bdev = dev->bdev;
-	bvm->bi_sector = bvm->bi_sector;
 
 	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
 }
diff --git a/drivers/md/dm-stripe.c b/drivers/md/dm-stripe.c
index 35c94ff24ad..a087bf2a8d6 100644
--- a/drivers/md/dm-stripe.c
+++ b/drivers/md/dm-stripe.c
@@ -26,14 +26,12 @@ struct stripe {
 struct stripe_c {
 	uint32_t stripes;
 	int stripes_shift;
-	sector_t stripes_mask;
 
 	/* The size of this target / num. stripes */
 	sector_t stripe_width;
 
-	/* stripe chunk size */
-	uint32_t chunk_shift;
-	sector_t chunk_mask;
+	uint32_t chunk_size;
+	int chunk_size_shift;
 
 	/* Needed for handling events */
 	struct dm_target *ti;
@@ -91,7 +89,7 @@ static int get_stripe(struct dm_target *ti, struct stripe_c *sc,
 
 /*
  * Construct a striped mapping.
- * <number of stripes> <chunk size (2^^n)> [<dev_path> <offset>]+
+ * <number of stripes> <chunk size> [<dev_path> <offset>]+
  */
 static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
@@ -99,7 +97,6 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	sector_t width;
 	uint32_t stripes;
 	uint32_t chunk_size;
-	char *end;
 	int r;
 	unsigned int i;
 
@@ -108,34 +105,23 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -EINVAL;
 	}
 
-	stripes = simple_strtoul(argv[0], &end, 10);
-	if (!stripes || *end) {
+	if (kstrtouint(argv[0], 10, &stripes) || !stripes) {
 		ti->error = "Invalid stripe count";
 		return -EINVAL;
 	}
 
-	chunk_size = simple_strtoul(argv[1], &end, 10);
-	if (*end) {
+	if (kstrtouint(argv[1], 10, &chunk_size) || !chunk_size) {
 		ti->error = "Invalid chunk_size";
 		return -EINVAL;
 	}
 
-	/*
-	 * chunk_size is a power of two
-	 */
-	if (!is_power_of_2(chunk_size) ||
-	    (chunk_size < (PAGE_SIZE >> SECTOR_SHIFT))) {
-		ti->error = "Invalid chunk size";
-		return -EINVAL;
-	}
-
-	if (ti->len & (chunk_size - 1)) {
+	width = ti->len;
+	if (sector_div(width, chunk_size)) {
 		ti->error = "Target length not divisible by "
 		    "chunk size";
 		return -EINVAL;
 	}
 
-	width = ti->len;
 	if (sector_div(width, stripes)) {
 		ti->error = "Target length not divisible by "
 		    "number of stripes";
@@ -167,17 +153,21 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	if (stripes & (stripes - 1))
 		sc->stripes_shift = -1;
-	else {
-		sc->stripes_shift = ffs(stripes) - 1;
-		sc->stripes_mask = ((sector_t) stripes) - 1;
-	}
+	else
+		sc->stripes_shift = __ffs(stripes);
+
+	r = dm_set_target_max_io_len(ti, chunk_size);
+	if (r)
+		return r;
 
-	ti->split_io = chunk_size;
 	ti->num_flush_requests = stripes;
 	ti->num_discard_requests = stripes;
 
-	sc->chunk_shift = ffs(chunk_size) - 1;
-	sc->chunk_mask = ((sector_t) chunk_size) - 1;
+	sc->chunk_size = chunk_size;
+	if (chunk_size & (chunk_size - 1))
+		sc->chunk_size_shift = -1;
+	else
+		sc->chunk_size_shift = __ffs(chunk_size);
 
 	/*
 	 * Get the stripe destinations.
@@ -216,17 +206,29 @@ static void stripe_dtr(struct dm_target *ti)
 static void stripe_map_sector(struct stripe_c *sc, sector_t sector,
 			      uint32_t *stripe, sector_t *result)
 {
-	sector_t offset = dm_target_offset(sc->ti, sector);
-	sector_t chunk = offset >> sc->chunk_shift;
+	sector_t chunk = dm_target_offset(sc->ti, sector);
+	sector_t chunk_offset;
+
+	if (sc->chunk_size_shift < 0)
+		chunk_offset = sector_div(chunk, sc->chunk_size);
+	else {
+		chunk_offset = chunk & (sc->chunk_size - 1);
+		chunk >>= sc->chunk_size_shift;
+	}
 
 	if (sc->stripes_shift < 0)
 		*stripe = sector_div(chunk, sc->stripes);
 	else {
-		*stripe = chunk & sc->stripes_mask;
+		*stripe = chunk & (sc->stripes - 1);
 		chunk >>= sc->stripes_shift;
 	}
 
-	*result = (chunk << sc->chunk_shift) | (offset & sc->chunk_mask);
+	if (sc->chunk_size_shift < 0)
+		chunk *= sc->chunk_size;
+	else
+		chunk <<= sc->chunk_size_shift;
+
+	*result = chunk + chunk_offset;
 }
 
 static void stripe_map_range_sector(struct stripe_c *sc, sector_t sector,
@@ -237,9 +239,16 @@ static void stripe_map_range_sector(struct stripe_c *sc, sector_t sector,
 	stripe_map_sector(sc, sector, &stripe, result);
 	if (stripe == target_stripe)
 		return;
-	*result &= ~sc->chunk_mask;			/* round down */
+
+	/* round down */
+	sector = *result;
+	if (sc->chunk_size_shift < 0)
+		*result -= sector_div(sector, sc->chunk_size);
+	else
+		*result = sector & ~(sector_t)(sc->chunk_size - 1);
+
 	if (target_stripe < stripe)
-		*result += sc->chunk_mask + 1;		/* next chunk */
+		*result += sc->chunk_size;		/* next chunk */
 }
 
 static int stripe_map_discard(struct stripe_c *sc, struct bio *bio,
@@ -302,8 +311,8 @@ static int stripe_map(struct dm_target *ti, struct bio *bio,
  *
  */
 
-static int stripe_status(struct dm_target *ti,
-			 status_type_t type, char *result, unsigned int maxlen)
+static int stripe_status(struct dm_target *ti, status_type_t type,
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct stripe_c *sc = (struct stripe_c *) ti->private;
 	char buffer[sc->stripes + 1];
@@ -324,7 +333,7 @@ static int stripe_status(struct dm_target *ti,
 
 	case STATUSTYPE_TABLE:
 		DMEMIT("%d %llu", sc->stripes,
-			(unsigned long long)sc->chunk_mask + 1);
+			(unsigned long long)sc->chunk_size);
 		for (i = 0; i < sc->stripes; i++)
 			DMEMIT(" %s %llu", sc->stripe[i].dev->name,
 			    (unsigned long long)sc->stripe[i].physical_start);
@@ -391,7 +400,7 @@ static void stripe_io_hints(struct dm_target *ti,
 			    struct queue_limits *limits)
 {
 	struct stripe_c *sc = ti->private;
-	unsigned chunk_size = (sc->chunk_mask + 1) << 9;
+	unsigned chunk_size = sc->chunk_size << SECTOR_SHIFT;
 
 	blk_limits_io_min(limits, chunk_size);
 	blk_limits_io_opt(limits, chunk_size * sc->stripes);
@@ -419,7 +428,7 @@ static int stripe_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
 
 static struct target_type stripe_target = {
 	.name   = "striped",
-	.version = {1, 4, 0},
+	.version = {1, 5, 0},
 	.module = THIS_MODULE,
 	.ctr    = stripe_ctr,
 	.dtr    = stripe_dtr,
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 2e227fbf162..f90069029aa 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -1319,6 +1319,9 @@ static bool dm_table_supports_flush(struct dm_table *t, unsigned flush)
 		if (!ti->num_flush_requests)
 			continue;
 
+		if (ti->flush_supported)
+			return 1;
+
 		if (ti->type->iterate_devices &&
 		    ti->type->iterate_devices(ti, device_flush_capable, &flush))
 			return 1;
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
index 3e2907f0bc4..693e149e972 100644
--- a/drivers/md/dm-thin-metadata.c
+++ b/drivers/md/dm-thin-metadata.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011 Red Hat, Inc.
+ * Copyright (C) 2011-2012 Red Hat, Inc.
  *
  * This file is released under the GPL.
  */
@@ -80,6 +80,12 @@
 #define THIN_METADATA_CACHE_SIZE 64
 #define SECTOR_TO_BLOCK_SHIFT 3
 
+/*
+ *  3 for btree insert +
+ *  2 for btree lookup used within space map
+ */
+#define THIN_MAX_CONCURRENT_LOCKS 5
+
 /* This should be plenty */
 #define SPACE_MAP_ROOT_SIZE 128
 
@@ -172,13 +178,20 @@ struct dm_pool_metadata {
 
 	struct rw_semaphore root_lock;
 	uint32_t time;
-	int need_commit;
 	dm_block_t root;
 	dm_block_t details_root;
 	struct list_head thin_devices;
 	uint64_t trans_id;
 	unsigned long flags;
 	sector_t data_block_size;
+	bool read_only:1;
+
+	/*
+	 * Set if a transaction has to be aborted but the attempt to roll back
+	 * to the previous (good) transaction failed.  The only pool metadata
+	 * operation possible in this state is the closing of the device.
+	 */
+	bool fail_io:1;
 };
 
 struct dm_thin_device {
@@ -187,7 +200,8 @@ struct dm_thin_device {
 	dm_thin_id id;
 
 	int open_count;
-	int changed;
+	bool changed:1;
+	bool aborted_with_changes:1;
 	uint64_t mapped_blocks;
 	uint64_t transaction_id;
 	uint32_t creation_time;
@@ -338,7 +352,21 @@ static int subtree_equal(void *context, void *value1_le, void *value2_le)
 
 /*----------------------------------------------------------------*/
 
-static int superblock_all_zeroes(struct dm_block_manager *bm, int *result)
+static int superblock_lock_zero(struct dm_pool_metadata *pmd,
+				struct dm_block **sblock)
+{
+	return dm_bm_write_lock_zero(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+				     &sb_validator, sblock);
+}
+
+static int superblock_lock(struct dm_pool_metadata *pmd,
+			   struct dm_block **sblock)
+{
+	return dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+				&sb_validator, sblock);
+}
+
+static int __superblock_all_zeroes(struct dm_block_manager *bm, int *result)
 {
 	int r;
 	unsigned i;
@@ -365,72 +393,9 @@ static int superblock_all_zeroes(struct dm_block_manager *bm, int *result)
 	return dm_bm_unlock(b);
 }
 
-static int init_pmd(struct dm_pool_metadata *pmd,
-		    struct dm_block_manager *bm,
-		    dm_block_t nr_blocks, int create)
+static void __setup_btree_details(struct dm_pool_metadata *pmd)
 {
-	int r;
-	struct dm_space_map *sm, *data_sm;
-	struct dm_transaction_manager *tm;
-	struct dm_block *sblock;
-
-	if (create) {
-		r = dm_tm_create_with_sm(bm, THIN_SUPERBLOCK_LOCATION,
-					 &sb_validator, &tm, &sm, &sblock);
-		if (r < 0) {
-			DMERR("tm_create_with_sm failed");
-			return r;
-		}
-
-		data_sm = dm_sm_disk_create(tm, nr_blocks);
-		if (IS_ERR(data_sm)) {
-			DMERR("sm_disk_create failed");
-			dm_tm_unlock(tm, sblock);
-			r = PTR_ERR(data_sm);
-			goto bad;
-		}
-	} else {
-		struct thin_disk_superblock *disk_super = NULL;
-		size_t space_map_root_offset =
-			offsetof(struct thin_disk_superblock, metadata_space_map_root);
-
-		r = dm_tm_open_with_sm(bm, THIN_SUPERBLOCK_LOCATION,
-				       &sb_validator, space_map_root_offset,
-				       SPACE_MAP_ROOT_SIZE, &tm, &sm, &sblock);
-		if (r < 0) {
-			DMERR("tm_open_with_sm failed");
-			return r;
-		}
-
-		disk_super = dm_block_data(sblock);
-		data_sm = dm_sm_disk_open(tm, disk_super->data_space_map_root,
-					  sizeof(disk_super->data_space_map_root));
-		if (IS_ERR(data_sm)) {
-			DMERR("sm_disk_open failed");
-			r = PTR_ERR(data_sm);
-			goto bad;
-		}
-	}
-
-
-	r = dm_tm_unlock(tm, sblock);
-	if (r < 0) {
-		DMERR("couldn't unlock superblock");
-		goto bad_data_sm;
-	}
-
-	pmd->bm = bm;
-	pmd->metadata_sm = sm;
-	pmd->data_sm = data_sm;
-	pmd->tm = tm;
-	pmd->nb_tm = dm_tm_create_non_blocking_clone(tm);
-	if (!pmd->nb_tm) {
-		DMERR("could not create clone tm");
-		r = -ENOMEM;
-		goto bad_data_sm;
-	}
-
-	pmd->info.tm = tm;
+	pmd->info.tm = pmd->tm;
 	pmd->info.levels = 2;
 	pmd->info.value_type.context = pmd->data_sm;
 	pmd->info.value_type.size = sizeof(__le64);
@@ -441,7 +406,7 @@ static int init_pmd(struct dm_pool_metadata *pmd,
 	memcpy(&pmd->nb_info, &pmd->info, sizeof(pmd->nb_info));
 	pmd->nb_info.tm = pmd->nb_tm;
 
-	pmd->tl_info.tm = tm;
+	pmd->tl_info.tm = pmd->tm;
 	pmd->tl_info.levels = 1;
 	pmd->tl_info.value_type.context = &pmd->info;
 	pmd->tl_info.value_type.size = sizeof(__le64);
@@ -449,7 +414,7 @@ static int init_pmd(struct dm_pool_metadata *pmd,
 	pmd->tl_info.value_type.dec = subtree_dec;
 	pmd->tl_info.value_type.equal = subtree_equal;
 
-	pmd->bl_info.tm = tm;
+	pmd->bl_info.tm = pmd->tm;
 	pmd->bl_info.levels = 1;
 	pmd->bl_info.value_type.context = pmd->data_sm;
 	pmd->bl_info.value_type.size = sizeof(__le64);
@@ -457,48 +422,266 @@ static int init_pmd(struct dm_pool_metadata *pmd,
 	pmd->bl_info.value_type.dec = data_block_dec;
 	pmd->bl_info.value_type.equal = data_block_equal;
 
-	pmd->details_info.tm = tm;
+	pmd->details_info.tm = pmd->tm;
 	pmd->details_info.levels = 1;
 	pmd->details_info.value_type.context = NULL;
 	pmd->details_info.value_type.size = sizeof(struct disk_device_details);
 	pmd->details_info.value_type.inc = NULL;
 	pmd->details_info.value_type.dec = NULL;
 	pmd->details_info.value_type.equal = NULL;
+}
 
-	pmd->root = 0;
+static int __write_initial_superblock(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_block *sblock;
+	size_t metadata_len, data_len;
+	struct thin_disk_superblock *disk_super;
+	sector_t bdev_size = i_size_read(pmd->bdev->bd_inode) >> SECTOR_SHIFT;
 
-	init_rwsem(&pmd->root_lock);
-	pmd->time = 0;
-	pmd->need_commit = 0;
-	pmd->details_root = 0;
-	pmd->trans_id = 0;
-	pmd->flags = 0;
-	INIT_LIST_HEAD(&pmd->thin_devices);
+	if (bdev_size > THIN_METADATA_MAX_SECTORS)
+		bdev_size = THIN_METADATA_MAX_SECTORS;
+
+	r = dm_sm_root_size(pmd->metadata_sm, &metadata_len);
+	if (r < 0)
+		return r;
+
+	r = dm_sm_root_size(pmd->data_sm, &data_len);
+	if (r < 0)
+		return r;
+
+	r = dm_sm_commit(pmd->data_sm);
+	if (r < 0)
+		return r;
+
+	r = dm_tm_pre_commit(pmd->tm);
+	if (r < 0)
+		return r;
+
+	r = superblock_lock_zero(pmd, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	disk_super->flags = 0;
+	memset(disk_super->uuid, 0, sizeof(disk_super->uuid));
+	disk_super->magic = cpu_to_le64(THIN_SUPERBLOCK_MAGIC);
+	disk_super->version = cpu_to_le32(THIN_VERSION);
+	disk_super->time = 0;
+	disk_super->trans_id = 0;
+	disk_super->held_root = 0;
+
+	r = dm_sm_copy_root(pmd->metadata_sm, &disk_super->metadata_space_map_root,
+			    metadata_len);
+	if (r < 0)
+		goto bad_locked;
+
+	r = dm_sm_copy_root(pmd->data_sm, &disk_super->data_space_map_root,
+			    data_len);
+	if (r < 0)
+		goto bad_locked;
+
+	disk_super->data_mapping_root = cpu_to_le64(pmd->root);
+	disk_super->device_details_root = cpu_to_le64(pmd->details_root);
+	disk_super->metadata_block_size = cpu_to_le32(THIN_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
+	disk_super->metadata_nr_blocks = cpu_to_le64(bdev_size >> SECTOR_TO_BLOCK_SHIFT);
+	disk_super->data_block_size = cpu_to_le32(pmd->data_block_size);
+
+	return dm_tm_commit(pmd->tm, sblock);
+
+bad_locked:
+	dm_bm_unlock(sblock);
+	return r;
+}
+
+static int __format_metadata(struct dm_pool_metadata *pmd)
+{
+	int r;
+
+	r = dm_tm_create_with_sm(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+				 &pmd->tm, &pmd->metadata_sm);
+	if (r < 0) {
+		DMERR("tm_create_with_sm failed");
+		return r;
+	}
+
+	pmd->data_sm = dm_sm_disk_create(pmd->tm, 0);
+	if (IS_ERR(pmd->data_sm)) {
+		DMERR("sm_disk_create failed");
+		r = PTR_ERR(pmd->data_sm);
+		goto bad_cleanup_tm;
+	}
+
+	pmd->nb_tm = dm_tm_create_non_blocking_clone(pmd->tm);
+	if (!pmd->nb_tm) {
+		DMERR("could not create non-blocking clone tm");
+		r = -ENOMEM;
+		goto bad_cleanup_data_sm;
+	}
+
+	__setup_btree_details(pmd);
+
+	r = dm_btree_empty(&pmd->info, &pmd->root);
+	if (r < 0)
+		goto bad_cleanup_nb_tm;
+
+	r = dm_btree_empty(&pmd->details_info, &pmd->details_root);
+	if (r < 0) {
+		DMERR("couldn't create devices root");
+		goto bad_cleanup_nb_tm;
+	}
+
+	r = __write_initial_superblock(pmd);
+	if (r)
+		goto bad_cleanup_nb_tm;
 
 	return 0;
 
-bad_data_sm:
-	dm_sm_destroy(data_sm);
-bad:
-	dm_tm_destroy(tm);
-	dm_sm_destroy(sm);
+bad_cleanup_nb_tm:
+	dm_tm_destroy(pmd->nb_tm);
+bad_cleanup_data_sm:
+	dm_sm_destroy(pmd->data_sm);
+bad_cleanup_tm:
+	dm_tm_destroy(pmd->tm);
+	dm_sm_destroy(pmd->metadata_sm);
+
+	return r;
+}
+
+static int __check_incompat_features(struct thin_disk_superblock *disk_super,
+				     struct dm_pool_metadata *pmd)
+{
+	uint32_t features;
+
+	features = le32_to_cpu(disk_super->incompat_flags) & ~THIN_FEATURE_INCOMPAT_SUPP;
+	if (features) {
+		DMERR("could not access metadata due to unsupported optional features (%lx).",
+		      (unsigned long)features);
+		return -EINVAL;
+	}
+
+	/*
+	 * Check for read-only metadata to skip the following RDWR checks.
+	 */
+	if (get_disk_ro(pmd->bdev->bd_disk))
+		return 0;
+
+	features = le32_to_cpu(disk_super->compat_ro_flags) & ~THIN_FEATURE_COMPAT_RO_SUPP;
+	if (features) {
+		DMERR("could not access metadata RDWR due to unsupported optional features (%lx).",
+		      (unsigned long)features);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int __open_metadata(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct thin_disk_superblock *disk_super;
+
+	r = dm_bm_read_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			    &sb_validator, &sblock);
+	if (r < 0) {
+		DMERR("couldn't read superblock");
+		return r;
+	}
+
+	disk_super = dm_block_data(sblock);
+
+	r = __check_incompat_features(disk_super, pmd);
+	if (r < 0)
+		goto bad_unlock_sblock;
+
+	r = dm_tm_open_with_sm(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			       disk_super->metadata_space_map_root,
+			       sizeof(disk_super->metadata_space_map_root),
+			       &pmd->tm, &pmd->metadata_sm);
+	if (r < 0) {
+		DMERR("tm_open_with_sm failed");
+		goto bad_unlock_sblock;
+	}
+
+	pmd->data_sm = dm_sm_disk_open(pmd->tm, disk_super->data_space_map_root,
+				       sizeof(disk_super->data_space_map_root));
+	if (IS_ERR(pmd->data_sm)) {
+		DMERR("sm_disk_open failed");
+		r = PTR_ERR(pmd->data_sm);
+		goto bad_cleanup_tm;
+	}
+
+	pmd->nb_tm = dm_tm_create_non_blocking_clone(pmd->tm);
+	if (!pmd->nb_tm) {
+		DMERR("could not create non-blocking clone tm");
+		r = -ENOMEM;
+		goto bad_cleanup_data_sm;
+	}
+
+	__setup_btree_details(pmd);
+	return dm_bm_unlock(sblock);
+
+bad_cleanup_data_sm:
+	dm_sm_destroy(pmd->data_sm);
+bad_cleanup_tm:
+	dm_tm_destroy(pmd->tm);
+	dm_sm_destroy(pmd->metadata_sm);
+bad_unlock_sblock:
+	dm_bm_unlock(sblock);
+
+	return r;
+}
+
+static int __open_or_format_metadata(struct dm_pool_metadata *pmd, bool format_device)
+{
+	int r, unformatted;
+
+	r = __superblock_all_zeroes(pmd->bm, &unformatted);
+	if (r)
+		return r;
+
+	if (unformatted)
+		return format_device ? __format_metadata(pmd) : -EPERM;
+
+	return __open_metadata(pmd);
+}
+
+static int __create_persistent_data_objects(struct dm_pool_metadata *pmd, bool format_device)
+{
+	int r;
+
+	pmd->bm = dm_block_manager_create(pmd->bdev, THIN_METADATA_BLOCK_SIZE,
+					  THIN_METADATA_CACHE_SIZE,
+					  THIN_MAX_CONCURRENT_LOCKS);
+	if (IS_ERR(pmd->bm)) {
+		DMERR("could not create block manager");
+		return PTR_ERR(pmd->bm);
+	}
+
+	r = __open_or_format_metadata(pmd, format_device);
+	if (r)
+		dm_block_manager_destroy(pmd->bm);
 
 	return r;
 }
 
+static void __destroy_persistent_data_objects(struct dm_pool_metadata *pmd)
+{
+	dm_sm_destroy(pmd->data_sm);
+	dm_sm_destroy(pmd->metadata_sm);
+	dm_tm_destroy(pmd->nb_tm);
+	dm_tm_destroy(pmd->tm);
+	dm_block_manager_destroy(pmd->bm);
+}
+
 static int __begin_transaction(struct dm_pool_metadata *pmd)
 {
 	int r;
-	u32 features;
 	struct thin_disk_superblock *disk_super;
 	struct dm_block *sblock;
 
 	/*
-	 * __maybe_commit_transaction() resets these
-	 */
-	WARN_ON(pmd->need_commit);
-
-	/*
 	 * We re-read the superblock every time.  Shouldn't need to do this
 	 * really.
 	 */
@@ -515,32 +698,8 @@ static int __begin_transaction(struct dm_pool_metadata *pmd)
 	pmd->flags = le32_to_cpu(disk_super->flags);
 	pmd->data_block_size = le32_to_cpu(disk_super->data_block_size);
 
-	features = le32_to_cpu(disk_super->incompat_flags) & ~THIN_FEATURE_INCOMPAT_SUPP;
-	if (features) {
-		DMERR("could not access metadata due to "
-		      "unsupported optional features (%lx).",
-		      (unsigned long)features);
-		r = -EINVAL;
-		goto out;
-	}
-
-	/*
-	 * Check for read-only metadata to skip the following RDWR checks.
-	 */
-	if (get_disk_ro(pmd->bdev->bd_disk))
-		goto out;
-
-	features = le32_to_cpu(disk_super->compat_ro_flags) & ~THIN_FEATURE_COMPAT_RO_SUPP;
-	if (features) {
-		DMERR("could not access metadata RDWR due to "
-		      "unsupported optional features (%lx).",
-		      (unsigned long)features);
-		r = -EINVAL;
-	}
-
-out:
 	dm_bm_unlock(sblock);
-	return r;
+	return 0;
 }
 
 static int __write_changed_details(struct dm_pool_metadata *pmd)
@@ -573,8 +732,6 @@ static int __write_changed_details(struct dm_pool_metadata *pmd)
 			list_del(&td->list);
 			kfree(td);
 		}
-
-		pmd->need_commit = 1;
 	}
 
 	return 0;
@@ -582,9 +739,6 @@ static int __write_changed_details(struct dm_pool_metadata *pmd)
 
 static int __commit_transaction(struct dm_pool_metadata *pmd)
 {
-	/*
-	 * FIXME: Associated pool should be made read-only on failure.
-	 */
 	int r;
 	size_t metadata_len, data_len;
 	struct thin_disk_superblock *disk_super;
@@ -597,31 +751,27 @@ static int __commit_transaction(struct dm_pool_metadata *pmd)
 
 	r = __write_changed_details(pmd);
 	if (r < 0)
-		goto out;
-
-	if (!pmd->need_commit)
-		goto out;
+		return r;
 
 	r = dm_sm_commit(pmd->data_sm);
 	if (r < 0)
-		goto out;
+		return r;
 
 	r = dm_tm_pre_commit(pmd->tm);
 	if (r < 0)
-		goto out;
+		return r;
 
 	r = dm_sm_root_size(pmd->metadata_sm, &metadata_len);
 	if (r < 0)
-		goto out;
+		return r;
 
 	r = dm_sm_root_size(pmd->data_sm, &data_len);
 	if (r < 0)
-		goto out;
+		return r;
 
-	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
-			     &sb_validator, &sblock);
+	r = superblock_lock(pmd, &sblock);
 	if (r)
-		goto out;
+		return r;
 
 	disk_super = dm_block_data(sblock);
 	disk_super->time = cpu_to_le32(pmd->time);
@@ -640,12 +790,7 @@ static int __commit_transaction(struct dm_pool_metadata *pmd)
 	if (r < 0)
 		goto out_locked;
 
-	r = dm_tm_commit(pmd->tm, sblock);
-	if (!r)
-		pmd->need_commit = 0;
-
-out:
-	return r;
+	return dm_tm_commit(pmd->tm, sblock);
 
 out_locked:
 	dm_bm_unlock(sblock);
@@ -653,15 +798,11 @@ out_locked:
 }
 
 struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
-					       sector_t data_block_size)
+					       sector_t data_block_size,
+					       bool format_device)
 {
 	int r;
-	struct thin_disk_superblock *disk_super;
 	struct dm_pool_metadata *pmd;
-	sector_t bdev_size = i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
-	struct dm_block_manager *bm;
-	int create;
-	struct dm_block *sblock;
 
 	pmd = kmalloc(sizeof(*pmd), GFP_KERNEL);
 	if (!pmd) {
@@ -669,90 +810,28 @@ struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
 		return ERR_PTR(-ENOMEM);
 	}
 
-	/*
-	 * Max hex locks:
-	 *  3 for btree insert +
-	 *  2 for btree lookup used within space map
-	 */
-	bm = dm_block_manager_create(bdev, THIN_METADATA_BLOCK_SIZE,
-				     THIN_METADATA_CACHE_SIZE, 5);
-	if (!bm) {
-		DMERR("could not create block manager");
-		kfree(pmd);
-		return ERR_PTR(-ENOMEM);
-	}
-
-	r = superblock_all_zeroes(bm, &create);
-	if (r) {
-		dm_block_manager_destroy(bm);
-		kfree(pmd);
-		return ERR_PTR(r);
-	}
-
+	init_rwsem(&pmd->root_lock);
+	pmd->time = 0;
+	INIT_LIST_HEAD(&pmd->thin_devices);
+	pmd->read_only = false;
+	pmd->fail_io = false;
+	pmd->bdev = bdev;
+	pmd->data_block_size = data_block_size;
 
-	r = init_pmd(pmd, bm, 0, create);
+	r = __create_persistent_data_objects(pmd, format_device);
 	if (r) {
-		dm_block_manager_destroy(bm);
 		kfree(pmd);
 		return ERR_PTR(r);
 	}
-	pmd->bdev = bdev;
-
-	if (!create) {
-		r = __begin_transaction(pmd);
-		if (r < 0)
-			goto bad;
-		return pmd;
-	}
-
-	/*
-	 * Create.
-	 */
-	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
-			     &sb_validator, &sblock);
-	if (r)
-		goto bad;
-
-	if (bdev_size > THIN_METADATA_MAX_SECTORS)
-		bdev_size = THIN_METADATA_MAX_SECTORS;
-
-	disk_super = dm_block_data(sblock);
-	disk_super->magic = cpu_to_le64(THIN_SUPERBLOCK_MAGIC);
-	disk_super->version = cpu_to_le32(THIN_VERSION);
-	disk_super->time = 0;
-	disk_super->metadata_block_size = cpu_to_le32(THIN_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
-	disk_super->metadata_nr_blocks = cpu_to_le64(bdev_size >> SECTOR_TO_BLOCK_SHIFT);
-	disk_super->data_block_size = cpu_to_le32(data_block_size);
-
-	r = dm_bm_unlock(sblock);
-	if (r < 0)
-		goto bad;
-
-	r = dm_btree_empty(&pmd->info, &pmd->root);
-	if (r < 0)
-		goto bad;
-
-	r = dm_btree_empty(&pmd->details_info, &pmd->details_root);
-	if (r < 0) {
-		DMERR("couldn't create devices root");
-		goto bad;
-	}
 
-	pmd->flags = 0;
-	pmd->need_commit = 1;
-	r = dm_pool_commit_metadata(pmd);
+	r = __begin_transaction(pmd);
 	if (r < 0) {
-		DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
-		      __func__, r);
-		goto bad;
+		if (dm_pool_metadata_close(pmd) < 0)
+			DMWARN("%s: dm_pool_metadata_close() failed.", __func__);
+		return ERR_PTR(r);
 	}
 
 	return pmd;
-
-bad:
-	if (dm_pool_metadata_close(pmd) < 0)
-		DMWARN("%s: dm_pool_metadata_close() failed.", __func__);
-	return ERR_PTR(r);
 }
 
 int dm_pool_metadata_close(struct dm_pool_metadata *pmd)
@@ -778,18 +857,17 @@ int dm_pool_metadata_close(struct dm_pool_metadata *pmd)
 		return -EBUSY;
 	}
 
-	r = __commit_transaction(pmd);
-	if (r < 0)
-		DMWARN("%s: __commit_transaction() failed, error = %d",
-		       __func__, r);
+	if (!pmd->read_only && !pmd->fail_io) {
+		r = __commit_transaction(pmd);
+		if (r < 0)
+			DMWARN("%s: __commit_transaction() failed, error = %d",
+			       __func__, r);
+	}
 
-	dm_tm_destroy(pmd->tm);
-	dm_tm_destroy(pmd->nb_tm);
-	dm_block_manager_destroy(pmd->bm);
-	dm_sm_destroy(pmd->metadata_sm);
-	dm_sm_destroy(pmd->data_sm);
-	kfree(pmd);
+	if (!pmd->fail_io)
+		__destroy_persistent_data_objects(pmd);
 
+	kfree(pmd);
 	return 0;
 }
 
@@ -850,6 +928,7 @@ static int __open_device(struct dm_pool_metadata *pmd,
 	(*td)->id = dev;
 	(*td)->open_count = 1;
 	(*td)->changed = changed;
+	(*td)->aborted_with_changes = false;
 	(*td)->mapped_blocks = le64_to_cpu(details_le.mapped_blocks);
 	(*td)->transaction_id = le64_to_cpu(details_le.transaction_id);
 	(*td)->creation_time = le32_to_cpu(details_le.creation_time);
@@ -911,10 +990,11 @@ static int __create_thin(struct dm_pool_metadata *pmd,
 
 int dm_pool_create_thin(struct dm_pool_metadata *pmd, dm_thin_id dev)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __create_thin(pmd, dev);
+	if (!pmd->fail_io)
+		r = __create_thin(pmd, dev);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1001,10 +1081,11 @@ int dm_pool_create_snap(struct dm_pool_metadata *pmd,
 				 dm_thin_id dev,
 				 dm_thin_id origin)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __create_snap(pmd, dev, origin);
+	if (!pmd->fail_io)
+		r = __create_snap(pmd, dev, origin);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1037,18 +1118,17 @@ static int __delete_device(struct dm_pool_metadata *pmd, dm_thin_id dev)
 	if (r)
 		return r;
 
-	pmd->need_commit = 1;
-
 	return 0;
 }
 
 int dm_pool_delete_thin_device(struct dm_pool_metadata *pmd,
 			       dm_thin_id dev)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __delete_device(pmd, dev);
+	if (!pmd->fail_io)
+		r = __delete_device(pmd, dev);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1058,28 +1138,40 @@ int dm_pool_set_metadata_transaction_id(struct dm_pool_metadata *pmd,
 					uint64_t current_id,
 					uint64_t new_id)
 {
+	int r = -EINVAL;
+
 	down_write(&pmd->root_lock);
+
+	if (pmd->fail_io)
+		goto out;
+
 	if (pmd->trans_id != current_id) {
-		up_write(&pmd->root_lock);
 		DMERR("mismatched transaction id");
-		return -EINVAL;
+		goto out;
 	}
 
 	pmd->trans_id = new_id;
-	pmd->need_commit = 1;
+	r = 0;
+
+out:
 	up_write(&pmd->root_lock);
 
-	return 0;
+	return r;
 }
 
 int dm_pool_get_metadata_transaction_id(struct dm_pool_metadata *pmd,
 					uint64_t *result)
 {
+	int r = -EINVAL;
+
 	down_read(&pmd->root_lock);
-	*result = pmd->trans_id;
+	if (!pmd->fail_io) {
+		*result = pmd->trans_id;
+		r = 0;
+	}
 	up_read(&pmd->root_lock);
 
-	return 0;
+	return r;
 }
 
 static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
@@ -1108,8 +1200,6 @@ static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
 
 		dm_tm_dec(pmd->tm, held_root);
 		dm_tm_unlock(pmd->tm, copy);
-		pmd->need_commit = 1;
-
 		return -EBUSY;
 	}
 
@@ -1131,29 +1221,25 @@ static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
 	/*
 	 * Write the held root into the superblock.
 	 */
-	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
-			     &sb_validator, &sblock);
+	r = superblock_lock(pmd, &sblock);
 	if (r) {
 		dm_tm_dec(pmd->tm, held_root);
-		pmd->need_commit = 1;
 		return r;
 	}
 
 	disk_super = dm_block_data(sblock);
 	disk_super->held_root = cpu_to_le64(held_root);
 	dm_bm_unlock(sblock);
-
-	pmd->need_commit = 1;
-
 	return 0;
 }
 
 int dm_pool_reserve_metadata_snap(struct dm_pool_metadata *pmd)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __reserve_metadata_snap(pmd);
+	if (!pmd->fail_io)
+		r = __reserve_metadata_snap(pmd);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1166,15 +1252,13 @@ static int __release_metadata_snap(struct dm_pool_metadata *pmd)
 	struct dm_block *sblock, *copy;
 	dm_block_t held_root;
 
-	r = dm_bm_write_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
-			     &sb_validator, &sblock);
+	r = superblock_lock(pmd, &sblock);
 	if (r)
 		return r;
 
 	disk_super = dm_block_data(sblock);
 	held_root = le64_to_cpu(disk_super->held_root);
 	disk_super->held_root = cpu_to_le64(0);
-	pmd->need_commit = 1;
 
 	dm_bm_unlock(sblock);
 
@@ -1197,10 +1281,11 @@ static int __release_metadata_snap(struct dm_pool_metadata *pmd)
 
 int dm_pool_release_metadata_snap(struct dm_pool_metadata *pmd)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __release_metadata_snap(pmd);
+	if (!pmd->fail_io)
+		r = __release_metadata_snap(pmd);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1227,10 +1312,11 @@ static int __get_metadata_snap(struct dm_pool_metadata *pmd,
 int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
 			      dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_read(&pmd->root_lock);
-	r = __get_metadata_snap(pmd, result);
+	if (!pmd->fail_io)
+		r = __get_metadata_snap(pmd, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1239,10 +1325,11 @@ int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
 int dm_pool_open_thin_device(struct dm_pool_metadata *pmd, dm_thin_id dev,
 			     struct dm_thin_device **td)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __open_device(pmd, dev, 0, td);
+	if (!pmd->fail_io)
+		r = __open_device(pmd, dev, 0, td);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1262,7 +1349,7 @@ dm_thin_id dm_thin_dev_id(struct dm_thin_device *td)
 	return td->id;
 }
 
-static int __snapshotted_since(struct dm_thin_device *td, uint32_t time)
+static bool __snapshotted_since(struct dm_thin_device *td, uint32_t time)
 {
 	return td->snapshotted_time > time;
 }
@@ -1270,28 +1357,31 @@ static int __snapshotted_since(struct dm_thin_device *td, uint32_t time)
 int dm_thin_find_block(struct dm_thin_device *td, dm_block_t block,
 		       int can_block, struct dm_thin_lookup_result *result)
 {
-	int r;
+	int r = -EINVAL;
 	uint64_t block_time = 0;
 	__le64 value;
 	struct dm_pool_metadata *pmd = td->pmd;
 	dm_block_t keys[2] = { td->id, block };
+	struct dm_btree_info *info;
 
 	if (can_block) {
 		down_read(&pmd->root_lock);
-		r = dm_btree_lookup(&pmd->info, pmd->root, keys, &value);
-		if (!r)
-			block_time = le64_to_cpu(value);
-		up_read(&pmd->root_lock);
-
-	} else if (down_read_trylock(&pmd->root_lock)) {
-		r = dm_btree_lookup(&pmd->nb_info, pmd->root, keys, &value);
-		if (!r)
-			block_time = le64_to_cpu(value);
-		up_read(&pmd->root_lock);
-
-	} else
+		info = &pmd->info;
+	} else if (down_read_trylock(&pmd->root_lock))
+		info = &pmd->nb_info;
+	else
 		return -EWOULDBLOCK;
 
+	if (pmd->fail_io)
+		goto out;
+
+	r = dm_btree_lookup(info, pmd->root, keys, &value);
+	if (!r)
+		block_time = le64_to_cpu(value);
+
+out:
+	up_read(&pmd->root_lock);
+
 	if (!r) {
 		dm_block_t exception_block;
 		uint32_t exception_time;
@@ -1312,7 +1402,6 @@ static int __insert(struct dm_thin_device *td, dm_block_t block,
 	struct dm_pool_metadata *pmd = td->pmd;
 	dm_block_t keys[2] = { td->id, block };
 
-	pmd->need_commit = 1;
 	value = cpu_to_le64(pack_block_time(data_block, pmd->time));
 	__dm_bless_for_disk(&value);
 
@@ -1321,10 +1410,9 @@ static int __insert(struct dm_thin_device *td, dm_block_t block,
 	if (r)
 		return r;
 
-	if (inserted) {
+	td->changed = 1;
+	if (inserted)
 		td->mapped_blocks++;
-		td->changed = 1;
-	}
 
 	return 0;
 }
@@ -1332,10 +1420,11 @@ static int __insert(struct dm_thin_device *td, dm_block_t block,
 int dm_thin_insert_block(struct dm_thin_device *td, dm_block_t block,
 			 dm_block_t data_block)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&td->pmd->root_lock);
-	r = __insert(td, block, data_block);
+	if (!td->pmd->fail_io)
+		r = __insert(td, block, data_block);
 	up_write(&td->pmd->root_lock);
 
 	return r;
@@ -1353,31 +1442,51 @@ static int __remove(struct dm_thin_device *td, dm_block_t block)
 
 	td->mapped_blocks--;
 	td->changed = 1;
-	pmd->need_commit = 1;
 
 	return 0;
 }
 
 int dm_thin_remove_block(struct dm_thin_device *td, dm_block_t block)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&td->pmd->root_lock);
-	r = __remove(td, block);
+	if (!td->pmd->fail_io)
+		r = __remove(td, block);
 	up_write(&td->pmd->root_lock);
 
 	return r;
 }
 
-int dm_pool_alloc_data_block(struct dm_pool_metadata *pmd, dm_block_t *result)
+bool dm_thin_changed_this_transaction(struct dm_thin_device *td)
 {
 	int r;
 
-	down_write(&pmd->root_lock);
+	down_read(&td->pmd->root_lock);
+	r = td->changed;
+	up_read(&td->pmd->root_lock);
 
-	r = dm_sm_new_block(pmd->data_sm, result);
-	pmd->need_commit = 1;
+	return r;
+}
+
+bool dm_thin_aborted_changes(struct dm_thin_device *td)
+{
+	bool r;
 
+	down_read(&td->pmd->root_lock);
+	r = td->aborted_with_changes;
+	up_read(&td->pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_alloc_data_block(struct dm_pool_metadata *pmd, dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = dm_sm_new_block(pmd->data_sm, result);
 	up_write(&pmd->root_lock);
 
 	return r;
@@ -1385,9 +1494,11 @@ int dm_pool_alloc_data_block(struct dm_pool_metadata *pmd, dm_block_t *result)
 
 int dm_pool_commit_metadata(struct dm_pool_metadata *pmd)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
+	if (pmd->fail_io)
+		goto out;
 
 	r = __commit_transaction(pmd);
 	if (r <= 0)
@@ -1402,12 +1513,41 @@ out:
 	return r;
 }
 
+static void __set_abort_with_changes_flags(struct dm_pool_metadata *pmd)
+{
+	struct dm_thin_device *td;
+
+	list_for_each_entry(td, &pmd->thin_devices, list)
+		td->aborted_with_changes = td->changed;
+}
+
+int dm_pool_abort_metadata(struct dm_pool_metadata *pmd)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (pmd->fail_io)
+		goto out;
+
+	__set_abort_with_changes_flags(pmd);
+	__destroy_persistent_data_objects(pmd);
+	r = __create_persistent_data_objects(pmd, false);
+	if (r)
+		pmd->fail_io = true;
+
+out:
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
 int dm_pool_get_free_block_count(struct dm_pool_metadata *pmd, dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_read(&pmd->root_lock);
-	r = dm_sm_get_nr_free(pmd->data_sm, result);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_free(pmd->data_sm, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1416,10 +1556,11 @@ int dm_pool_get_free_block_count(struct dm_pool_metadata *pmd, dm_block_t *resul
 int dm_pool_get_free_metadata_block_count(struct dm_pool_metadata *pmd,
 					  dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_read(&pmd->root_lock);
-	r = dm_sm_get_nr_free(pmd->metadata_sm, result);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_free(pmd->metadata_sm, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1428,10 +1569,11 @@ int dm_pool_get_free_metadata_block_count(struct dm_pool_metadata *pmd,
 int dm_pool_get_metadata_dev_size(struct dm_pool_metadata *pmd,
 				  dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_read(&pmd->root_lock);
-	r = dm_sm_get_nr_blocks(pmd->metadata_sm, result);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_blocks(pmd->metadata_sm, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1448,10 +1590,11 @@ int dm_pool_get_data_block_size(struct dm_pool_metadata *pmd, sector_t *result)
 
 int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_read(&pmd->root_lock);
-	r = dm_sm_get_nr_blocks(pmd->data_sm, result);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_blocks(pmd->data_sm, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1459,13 +1602,17 @@ int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result)
 
 int dm_thin_get_mapped_count(struct dm_thin_device *td, dm_block_t *result)
 {
+	int r = -EINVAL;
 	struct dm_pool_metadata *pmd = td->pmd;
 
 	down_read(&pmd->root_lock);
-	*result = td->mapped_blocks;
+	if (!pmd->fail_io) {
+		*result = td->mapped_blocks;
+		r = 0;
+	}
 	up_read(&pmd->root_lock);
 
-	return 0;
+	return r;
 }
 
 static int __highest_block(struct dm_thin_device *td, dm_block_t *result)
@@ -1487,11 +1634,12 @@ static int __highest_block(struct dm_thin_device *td, dm_block_t *result)
 int dm_thin_get_highest_mapped_block(struct dm_thin_device *td,
 				     dm_block_t *result)
 {
-	int r;
+	int r = -EINVAL;
 	struct dm_pool_metadata *pmd = td->pmd;
 
 	down_read(&pmd->root_lock);
-	r = __highest_block(td, result);
+	if (!pmd->fail_io)
+		r = __highest_block(td, result);
 	up_read(&pmd->root_lock);
 
 	return r;
@@ -1514,20 +1662,25 @@ static int __resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
 		return -EINVAL;
 	}
 
-	r = dm_sm_extend(pmd->data_sm, new_count - old_count);
-	if (!r)
-		pmd->need_commit = 1;
-
-	return r;
+	return dm_sm_extend(pmd->data_sm, new_count - old_count);
 }
 
 int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
 {
-	int r;
+	int r = -EINVAL;
 
 	down_write(&pmd->root_lock);
-	r = __resize_data_dev(pmd, new_count);
+	if (!pmd->fail_io)
+		r = __resize_data_dev(pmd, new_count);
 	up_write(&pmd->root_lock);
 
 	return r;
 }
+
+void dm_pool_metadata_read_only(struct dm_pool_metadata *pmd)
+{
+	down_write(&pmd->root_lock);
+	pmd->read_only = true;
+	dm_bm_set_read_only(pmd->bm);
+	up_write(&pmd->root_lock);
+}
diff --git a/drivers/md/dm-thin-metadata.h b/drivers/md/dm-thin-metadata.h
index b88918ccdaf..0cecc370288 100644
--- a/drivers/md/dm-thin-metadata.h
+++ b/drivers/md/dm-thin-metadata.h
@@ -38,7 +38,8 @@ typedef uint64_t dm_thin_id;
  * Reopens or creates a new, empty metadata volume.
  */
 struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
-					       sector_t data_block_size);
+					       sector_t data_block_size,
+					       bool format_device);
 
 int dm_pool_metadata_close(struct dm_pool_metadata *pmd);
 
@@ -79,6 +80,16 @@ int dm_pool_delete_thin_device(struct dm_pool_metadata *pmd,
 int dm_pool_commit_metadata(struct dm_pool_metadata *pmd);
 
 /*
+ * Discards all uncommitted changes.  Rereads the superblock, rolling back
+ * to the last good transaction.  Thin devices remain open.
+ * dm_thin_aborted_changes() tells you if they had uncommitted changes.
+ *
+ * If this call fails it's only useful to call dm_pool_metadata_close().
+ * All other methods will fail with -EINVAL.
+ */
+int dm_pool_abort_metadata(struct dm_pool_metadata *pmd);
+
+/*
  * Set/get userspace transaction id.
  */
 int dm_pool_set_metadata_transaction_id(struct dm_pool_metadata *pmd,
@@ -119,7 +130,7 @@ dm_thin_id dm_thin_dev_id(struct dm_thin_device *td);
 
 struct dm_thin_lookup_result {
 	dm_block_t block;
-	int shared;
+	unsigned shared:1;
 };
 
 /*
@@ -147,6 +158,10 @@ int dm_thin_remove_block(struct dm_thin_device *td, dm_block_t block);
 /*
  * Queries.
  */
+bool dm_thin_changed_this_transaction(struct dm_thin_device *td);
+
+bool dm_thin_aborted_changes(struct dm_thin_device *td);
+
 int dm_thin_get_highest_mapped_block(struct dm_thin_device *td,
 				     dm_block_t *highest_mapped);
 
@@ -171,6 +186,12 @@ int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result);
  */
 int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_size);
 
+/*
+ * Flicks the underlying block manager into read only mode, so you know
+ * that nothing is changing.
+ */
+void dm_pool_metadata_read_only(struct dm_pool_metadata *pmd);
+
 /*----------------------------------------------------------------*/
 
 #endif
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
index 37fdaf81bd1..af1fc3b2c2a 100644
--- a/drivers/md/dm-thin.c
+++ b/drivers/md/dm-thin.c
@@ -1,10 +1,11 @@
 /*
- * Copyright (C) 2011 Red Hat UK.
+ * Copyright (C) 2011-2012 Red Hat UK.
  *
  * This file is released under the GPL.
  */
 
 #include "dm-thin-metadata.h"
+#include "dm.h"
 
 #include <linux/device-mapper.h>
 #include <linux/dm-io.h>
@@ -19,7 +20,7 @@
 /*
  * Tunable constants
  */
-#define ENDIO_HOOK_POOL_SIZE 10240
+#define ENDIO_HOOK_POOL_SIZE 1024
 #define DEFERRED_SET_SIZE 64
 #define MAPPING_POOL_SIZE 1024
 #define PRISON_CELLS 1024
@@ -496,12 +497,27 @@ static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,
  */
 struct dm_thin_new_mapping;
 
+/*
+ * The pool runs in 3 modes.  Ordered in degraded order for comparisons.
+ */
+enum pool_mode {
+	PM_WRITE,		/* metadata may be changed */
+	PM_READ_ONLY,		/* metadata may not be changed */
+	PM_FAIL,		/* all I/O fails */
+};
+
 struct pool_features {
+	enum pool_mode mode;
+
 	unsigned zero_new_blocks:1;
 	unsigned discard_enabled:1;
 	unsigned discard_passdown:1;
 };
 
+struct thin_c;
+typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
+typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);
+
 struct pool {
 	struct list_head list;
 	struct dm_target *ti;	/* Only set if a pool target is bound */
@@ -510,10 +526,9 @@ struct pool {
 	struct block_device *md_dev;
 	struct dm_pool_metadata *pmd;
 
-	uint32_t sectors_per_block;
-	unsigned block_shift;
-	dm_block_t offset_mask;
 	dm_block_t low_water_blocks;
+	uint32_t sectors_per_block;
+	int sectors_per_block_shift;
 
 	struct pool_features pf;
 	unsigned low_water_triggered:1;	/* A dm event has been sent */
@@ -526,8 +541,8 @@ struct pool {
 	struct work_struct worker;
 	struct delayed_work waker;
 
-	unsigned ref_count;
 	unsigned long last_commit_jiffies;
+	unsigned ref_count;
 
 	spinlock_t lock;
 	struct bio_list deferred_bios;
@@ -543,8 +558,17 @@ struct pool {
 	struct dm_thin_new_mapping *next_mapping;
 	mempool_t *mapping_pool;
 	mempool_t *endio_hook_pool;
+
+	process_bio_fn process_bio;
+	process_bio_fn process_discard;
+
+	process_mapping_fn process_prepared_mapping;
+	process_mapping_fn process_prepared_discard;
 };
 
+static enum pool_mode get_pool_mode(struct pool *pool);
+static void set_pool_mode(struct pool *pool, enum pool_mode mode);
+
 /*
  * Target context for a pool.
  */
@@ -679,16 +703,28 @@ static void requeue_io(struct thin_c *tc)
 
 static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
 {
-	return bio->bi_sector >> tc->pool->block_shift;
+	sector_t block_nr = bio->bi_sector;
+
+	if (tc->pool->sectors_per_block_shift < 0)
+		(void) sector_div(block_nr, tc->pool->sectors_per_block);
+	else
+		block_nr >>= tc->pool->sectors_per_block_shift;
+
+	return block_nr;
 }
 
 static void remap(struct thin_c *tc, struct bio *bio, dm_block_t block)
 {
 	struct pool *pool = tc->pool;
+	sector_t bi_sector = bio->bi_sector;
 
 	bio->bi_bdev = tc->pool_dev->bdev;
-	bio->bi_sector = (block << pool->block_shift) +
-		(bio->bi_sector & pool->offset_mask);
+	if (tc->pool->sectors_per_block_shift < 0)
+		bio->bi_sector = (block * pool->sectors_per_block) +
+				 sector_div(bi_sector, pool->sectors_per_block);
+	else
+		bio->bi_sector = (block << pool->sectors_per_block_shift) |
+				(bi_sector & (pool->sectors_per_block - 1));
 }
 
 static void remap_to_origin(struct thin_c *tc, struct bio *bio)
@@ -696,21 +732,39 @@ static void remap_to_origin(struct thin_c *tc, struct bio *bio)
 	bio->bi_bdev = tc->origin_dev->bdev;
 }
 
+static int bio_triggers_commit(struct thin_c *tc, struct bio *bio)
+{
+	return (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) &&
+		dm_thin_changed_this_transaction(tc->td);
+}
+
 static void issue(struct thin_c *tc, struct bio *bio)
 {
 	struct pool *pool = tc->pool;
 	unsigned long flags;
 
+	if (!bio_triggers_commit(tc, bio)) {
+		generic_make_request(bio);
+		return;
+	}
+
 	/*
-	 * Batch together any FUA/FLUSH bios we find and then issue
-	 * a single commit for them in process_deferred_bios().
+	 * Complete bio with an error if earlier I/O caused changes to
+	 * the metadata that can't be committed e.g, due to I/O errors
+	 * on the metadata device.
 	 */
-	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
-		spin_lock_irqsave(&pool->lock, flags);
-		bio_list_add(&pool->deferred_flush_bios, bio);
-		spin_unlock_irqrestore(&pool->lock, flags);
-	} else
-		generic_make_request(bio);
+	if (dm_thin_aborted_changes(tc->td)) {
+		bio_io_error(bio);
+		return;
+	}
+
+	/*
+	 * Batch together any bios that trigger commits and then issue a
+	 * single commit for them in process_deferred_bios().
+	 */
+	spin_lock_irqsave(&pool->lock, flags);
+	bio_list_add(&pool->deferred_flush_bios, bio);
+	spin_unlock_irqrestore(&pool->lock, flags);
 }
 
 static void remap_to_origin_and_issue(struct thin_c *tc, struct bio *bio)
@@ -847,6 +901,14 @@ static void cell_defer_except(struct thin_c *tc, struct dm_bio_prison_cell *cell
 	wake_worker(pool);
 }
 
+static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
+{
+	if (m->bio)
+		m->bio->bi_end_io = m->saved_bi_end_io;
+	cell_error(m->cell);
+	list_del(&m->list);
+	mempool_free(m, m->tc->pool->mapping_pool);
+}
 static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 {
 	struct thin_c *tc = m->tc;
@@ -859,7 +921,7 @@ static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 
 	if (m->err) {
 		cell_error(m->cell);
-		return;
+		goto out;
 	}
 
 	/*
@@ -871,7 +933,7 @@ static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 	if (r) {
 		DMERR("dm_thin_insert_block() failed");
 		cell_error(m->cell);
-		return;
+		goto out;
 	}
 
 	/*
@@ -886,22 +948,25 @@ static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 	} else
 		cell_defer(tc, m->cell, m->data_block);
 
+out:
 	list_del(&m->list);
 	mempool_free(m, tc->pool->mapping_pool);
 }
 
-static void process_prepared_discard(struct dm_thin_new_mapping *m)
+static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
 {
-	int r;
 	struct thin_c *tc = m->tc;
 
-	r = dm_thin_remove_block(tc->td, m->virt_block);
-	if (r)
-		DMERR("dm_thin_remove_block() failed");
+	bio_io_error(m->bio);
+	cell_defer_except(tc, m->cell);
+	cell_defer_except(tc, m->cell2);
+	mempool_free(m, tc->pool->mapping_pool);
+}
+
+static void process_prepared_discard_passdown(struct dm_thin_new_mapping *m)
+{
+	struct thin_c *tc = m->tc;
 
-	/*
-	 * Pass the discard down to the underlying device?
-	 */
 	if (m->pass_discard)
 		remap_and_issue(tc, m->bio, m->data_block);
 	else
@@ -912,8 +977,20 @@ static void process_prepared_discard(struct dm_thin_new_mapping *m)
 	mempool_free(m, tc->pool->mapping_pool);
 }
 
+static void process_prepared_discard(struct dm_thin_new_mapping *m)
+{
+	int r;
+	struct thin_c *tc = m->tc;
+
+	r = dm_thin_remove_block(tc->td, m->virt_block);
+	if (r)
+		DMERR("dm_thin_remove_block() failed");
+
+	process_prepared_discard_passdown(m);
+}
+
 static void process_prepared(struct pool *pool, struct list_head *head,
-			     void (*fn)(struct dm_thin_new_mapping *))
+			     process_mapping_fn *fn)
 {
 	unsigned long flags;
 	struct list_head maps;
@@ -925,7 +1002,7 @@ static void process_prepared(struct pool *pool, struct list_head *head,
 	spin_unlock_irqrestore(&pool->lock, flags);
 
 	list_for_each_entry_safe(m, tmp, &maps, list)
-		fn(m);
+		(*fn)(m);
 }
 
 /*
@@ -933,9 +1010,7 @@ static void process_prepared(struct pool *pool, struct list_head *head,
  */
 static int io_overlaps_block(struct pool *pool, struct bio *bio)
 {
-	return !(bio->bi_sector & pool->offset_mask) &&
-		(bio->bi_size == (pool->sectors_per_block << SECTOR_SHIFT));
-
+	return bio->bi_size == (pool->sectors_per_block << SECTOR_SHIFT);
 }
 
 static int io_overwrites_block(struct pool *pool, struct bio *bio)
@@ -1093,6 +1168,35 @@ static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
 	}
 }
 
+static int commit(struct pool *pool)
+{
+	int r;
+
+	r = dm_pool_commit_metadata(pool->pmd);
+	if (r)
+		DMERR("commit failed, error = %d", r);
+
+	return r;
+}
+
+/*
+ * A non-zero return indicates read_only or fail_io mode.
+ * Many callers don't care about the return value.
+ */
+static int commit_or_fallback(struct pool *pool)
+{
+	int r;
+
+	if (get_pool_mode(pool) != PM_WRITE)
+		return -EINVAL;
+
+	r = commit(pool);
+	if (r)
+		set_pool_mode(pool, PM_READ_ONLY);
+
+	return r;
+}
+
 static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
 {
 	int r;
@@ -1121,12 +1225,7 @@ static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
 			 * Try to commit to see if that will free up some
 			 * more space.
 			 */
-			r = dm_pool_commit_metadata(pool->pmd);
-			if (r) {
-				DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
-				      __func__, r);
-				return r;
-			}
+			(void) commit_or_fallback(pool);
 
 			r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
 			if (r)
@@ -1218,7 +1317,7 @@ static void process_discard(struct thin_c *tc, struct bio *bio)
 			 */
 			m = get_next_mapping(pool);
 			m->tc = tc;
-			m->pass_discard = (!lookup_result.shared) & pool->pf.discard_passdown;
+			m->pass_discard = (!lookup_result.shared) && pool->pf.discard_passdown;
 			m->virt_block = block;
 			m->data_block = lookup_result.block;
 			m->cell = cell;
@@ -1234,18 +1333,16 @@ static void process_discard(struct thin_c *tc, struct bio *bio)
 			}
 		} else {
 			/*
-			 * This path is hit if people are ignoring
-			 * limits->discard_granularity.  It ignores any
-			 * part of the discard that is in a subsequent
-			 * block.
+			 * The DM core makes sure that the discard doesn't span
+			 * a block boundary.  So we submit the discard of a
+			 * partial block appropriately.
 			 */
-			sector_t offset = bio->bi_sector - (block << pool->block_shift);
-			unsigned remaining = (pool->sectors_per_block - offset) << 9;
-			bio->bi_size = min(bio->bi_size, remaining);
-
 			cell_release_singleton(cell, bio);
 			cell_release_singleton(cell2, bio);
-			remap_and_issue(tc, bio, lookup_result.block);
+			if ((!lookup_result.shared) && pool->pf.discard_passdown)
+				remap_and_issue(tc, bio, lookup_result.block);
+			else
+				bio_endio(bio, 0);
 		}
 		break;
 
@@ -1307,7 +1404,7 @@ static void process_shared_bio(struct thin_c *tc, struct bio *bio,
 	if (bio_detain(pool->prison, &key, bio, &cell))
 		return;
 
-	if (bio_data_dir(bio) == WRITE)
+	if (bio_data_dir(bio) == WRITE && bio->bi_size)
 		break_sharing(tc, bio, block, &key, lookup_result, cell);
 	else {
 		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;
@@ -1359,6 +1456,7 @@ static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block
 
 	default:
 		DMERR("%s: alloc_data_block() failed, error = %d", __func__, r);
+		set_pool_mode(tc->pool, PM_READ_ONLY);
 		cell_error(cell);
 		break;
 	}
@@ -1416,6 +1514,49 @@ static void process_bio(struct thin_c *tc, struct bio *bio)
 	}
 }
 
+static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
+{
+	int r;
+	int rw = bio_data_dir(bio);
+	dm_block_t block = get_bio_block(tc, bio);
+	struct dm_thin_lookup_result lookup_result;
+
+	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
+	switch (r) {
+	case 0:
+		if (lookup_result.shared && (rw == WRITE) && bio->bi_size)
+			bio_io_error(bio);
+		else
+			remap_and_issue(tc, bio, lookup_result.block);
+		break;
+
+	case -ENODATA:
+		if (rw != READ) {
+			bio_io_error(bio);
+			break;
+		}
+
+		if (tc->origin_dev) {
+			remap_to_origin_and_issue(tc, bio);
+			break;
+		}
+
+		zero_fill_bio(bio);
+		bio_endio(bio, 0);
+		break;
+
+	default:
+		DMERR("dm_thin_find_block() failed, error = %d", r);
+		bio_io_error(bio);
+		break;
+	}
+}
+
+static void process_bio_fail(struct thin_c *tc, struct bio *bio)
+{
+	bio_io_error(bio);
+}
+
 static int need_commit_due_to_time(struct pool *pool)
 {
 	return jiffies < pool->last_commit_jiffies ||
@@ -1427,7 +1568,6 @@ static void process_deferred_bios(struct pool *pool)
 	unsigned long flags;
 	struct bio *bio;
 	struct bio_list bios;
-	int r;
 
 	bio_list_init(&bios);
 
@@ -1454,9 +1594,9 @@ static void process_deferred_bios(struct pool *pool)
 		}
 
 		if (bio->bi_rw & REQ_DISCARD)
-			process_discard(tc, bio);
+			pool->process_discard(tc, bio);
 		else
-			process_bio(tc, bio);
+			pool->process_bio(tc, bio);
 	}
 
 	/*
@@ -1472,10 +1612,7 @@ static void process_deferred_bios(struct pool *pool)
 	if (bio_list_empty(&bios) && !need_commit_due_to_time(pool))
 		return;
 
-	r = dm_pool_commit_metadata(pool->pmd);
-	if (r) {
-		DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
-		      __func__, r);
+	if (commit_or_fallback(pool)) {
 		while ((bio = bio_list_pop(&bios)))
 			bio_io_error(bio);
 		return;
@@ -1490,8 +1627,8 @@ static void do_worker(struct work_struct *ws)
 {
 	struct pool *pool = container_of(ws, struct pool, worker);
 
-	process_prepared(pool, &pool->prepared_mappings, process_prepared_mapping);
-	process_prepared(pool, &pool->prepared_discards, process_prepared_discard);
+	process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
+	process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
 	process_deferred_bios(pool);
 }
 
@@ -1508,6 +1645,52 @@ static void do_waker(struct work_struct *ws)
 
 /*----------------------------------------------------------------*/
 
+static enum pool_mode get_pool_mode(struct pool *pool)
+{
+	return pool->pf.mode;
+}
+
+static void set_pool_mode(struct pool *pool, enum pool_mode mode)
+{
+	int r;
+
+	pool->pf.mode = mode;
+
+	switch (mode) {
+	case PM_FAIL:
+		DMERR("switching pool to failure mode");
+		pool->process_bio = process_bio_fail;
+		pool->process_discard = process_bio_fail;
+		pool->process_prepared_mapping = process_prepared_mapping_fail;
+		pool->process_prepared_discard = process_prepared_discard_fail;
+		break;
+
+	case PM_READ_ONLY:
+		DMERR("switching pool to read-only mode");
+		r = dm_pool_abort_metadata(pool->pmd);
+		if (r) {
+			DMERR("aborting transaction failed");
+			set_pool_mode(pool, PM_FAIL);
+		} else {
+			dm_pool_metadata_read_only(pool->pmd);
+			pool->process_bio = process_bio_read_only;
+			pool->process_discard = process_discard;
+			pool->process_prepared_mapping = process_prepared_mapping_fail;
+			pool->process_prepared_discard = process_prepared_discard_passdown;
+		}
+		break;
+
+	case PM_WRITE:
+		pool->process_bio = process_bio;
+		pool->process_discard = process_discard;
+		pool->process_prepared_mapping = process_prepared_mapping;
+		pool->process_prepared_discard = process_prepared_discard;
+		break;
+	}
+}
+
+/*----------------------------------------------------------------*/
+
 /*
  * Mapping functions.
  */
@@ -1553,6 +1736,12 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio,
 	struct dm_thin_lookup_result result;
 
 	map_context->ptr = thin_hook_bio(tc, bio);
+
+	if (get_pool_mode(tc->pool) == PM_FAIL) {
+		bio_io_error(bio);
+		return DM_MAPIO_SUBMITTED;
+	}
+
 	if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)) {
 		thin_defer_bio(tc, bio);
 		return DM_MAPIO_SUBMITTED;
@@ -1589,14 +1778,35 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio,
 		break;
 
 	case -ENODATA:
+		if (get_pool_mode(tc->pool) == PM_READ_ONLY) {
+			/*
+			 * This block isn't provisioned, and we have no way
+			 * of doing so.  Just error it.
+			 */
+			bio_io_error(bio);
+			r = DM_MAPIO_SUBMITTED;
+			break;
+		}
+		/* fall through */
+
+	case -EWOULDBLOCK:
 		/*
 		 * In future, the failed dm_thin_find_block above could
 		 * provide the hint to load the metadata into cache.
 		 */
-	case -EWOULDBLOCK:
 		thin_defer_bio(tc, bio);
 		r = DM_MAPIO_SUBMITTED;
 		break;
+
+	default:
+		/*
+		 * Must always call bio_io_error on failure.
+		 * dm_thin_find_block can fail with -EINVAL if the
+		 * pool is switched to fail-io mode.
+		 */
+		bio_io_error(bio);
+		r = DM_MAPIO_SUBMITTED;
+		break;
 	}
 
 	return r;
@@ -1633,15 +1843,26 @@ static int bind_control_target(struct pool *pool, struct dm_target *ti)
 {
 	struct pool_c *pt = ti->private;
 
+	/*
+	 * We want to make sure that degraded pools are never upgraded.
+	 */
+	enum pool_mode old_mode = pool->pf.mode;
+	enum pool_mode new_mode = pt->pf.mode;
+
+	if (old_mode > new_mode)
+		new_mode = old_mode;
+
 	pool->ti = ti;
 	pool->low_water_blocks = pt->low_water_blocks;
 	pool->pf = pt->pf;
+	set_pool_mode(pool, new_mode);
 
 	/*
 	 * If discard_passdown was enabled verify that the data device
 	 * supports discards.  Disable discard_passdown if not; otherwise
 	 * -EOPNOTSUPP will be returned.
 	 */
+	/* FIXME: pull this out into a sep fn. */
 	if (pt->pf.discard_passdown) {
 		struct request_queue *q = bdev_get_queue(pt->data_dev->bdev);
 		if (!q || !blk_queue_discard(q)) {
@@ -1667,6 +1888,7 @@ static void unbind_control_target(struct pool *pool, struct dm_target *ti)
 /* Initialize pool features. */
 static void pool_features_init(struct pool_features *pf)
 {
+	pf->mode = PM_WRITE;
 	pf->zero_new_blocks = 1;
 	pf->discard_enabled = 1;
 	pf->discard_passdown = 1;
@@ -1697,14 +1919,16 @@ static struct kmem_cache *_endio_hook_cache;
 
 static struct pool *pool_create(struct mapped_device *pool_md,
 				struct block_device *metadata_dev,
-				unsigned long block_size, char **error)
+				unsigned long block_size,
+				int read_only, char **error)
 {
 	int r;
 	void *err_p;
 	struct pool *pool;
 	struct dm_pool_metadata *pmd;
+	bool format_device = read_only ? false : true;
 
-	pmd = dm_pool_metadata_open(metadata_dev, block_size);
+	pmd = dm_pool_metadata_open(metadata_dev, block_size, format_device);
 	if (IS_ERR(pmd)) {
 		*error = "Error creating metadata object";
 		return (struct pool *)pmd;
@@ -1719,8 +1943,10 @@ static struct pool *pool_create(struct mapped_device *pool_md,
 
 	pool->pmd = pmd;
 	pool->sectors_per_block = block_size;
-	pool->block_shift = ffs(block_size) - 1;
-	pool->offset_mask = block_size - 1;
+	if (block_size & (block_size - 1))
+		pool->sectors_per_block_shift = -1;
+	else
+		pool->sectors_per_block_shift = __ffs(block_size);
 	pool->low_water_blocks = 0;
 	pool_features_init(&pool->pf);
 	pool->prison = prison_create(PRISON_CELLS);
@@ -1819,25 +2045,29 @@ static void __pool_dec(struct pool *pool)
 
 static struct pool *__pool_find(struct mapped_device *pool_md,
 				struct block_device *metadata_dev,
-				unsigned long block_size, char **error,
-				int *created)
+				unsigned long block_size, int read_only,
+				char **error, int *created)
 {
 	struct pool *pool = __pool_table_lookup_metadata_dev(metadata_dev);
 
 	if (pool) {
-		if (pool->pool_md != pool_md)
+		if (pool->pool_md != pool_md) {
+			*error = "metadata device already in use by a pool";
 			return ERR_PTR(-EBUSY);
+		}
 		__pool_inc(pool);
 
 	} else {
 		pool = __pool_table_lookup(pool_md);
 		if (pool) {
-			if (pool->md_dev != metadata_dev)
+			if (pool->md_dev != metadata_dev) {
+				*error = "different pool cannot replace a pool";
 				return ERR_PTR(-EINVAL);
+			}
 			__pool_inc(pool);
 
 		} else {
-			pool = pool_create(pool_md, metadata_dev, block_size, error);
+			pool = pool_create(pool_md, metadata_dev, block_size, read_only, error);
 			*created = 1;
 		}
 	}
@@ -1888,19 +2118,23 @@ static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
 		arg_name = dm_shift_arg(as);
 		argc--;
 
-		if (!strcasecmp(arg_name, "skip_block_zeroing")) {
+		if (!strcasecmp(arg_name, "skip_block_zeroing"))
 			pf->zero_new_blocks = 0;
-			continue;
-		} else if (!strcasecmp(arg_name, "ignore_discard")) {
+
+		else if (!strcasecmp(arg_name, "ignore_discard"))
 			pf->discard_enabled = 0;
-			continue;
-		} else if (!strcasecmp(arg_name, "no_discard_passdown")) {
+
+		else if (!strcasecmp(arg_name, "no_discard_passdown"))
 			pf->discard_passdown = 0;
-			continue;
-		}
 
-		ti->error = "Unrecognised pool feature requested";
-		r = -EINVAL;
+		else if (!strcasecmp(arg_name, "read_only"))
+			pf->mode = PM_READ_ONLY;
+
+		else {
+			ti->error = "Unrecognised pool feature requested";
+			r = -EINVAL;
+			break;
+		}
 	}
 
 	return r;
@@ -1964,7 +2198,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 	if (kstrtoul(argv[2], 10, &block_size) || !block_size ||
 	    block_size < DATA_DEV_BLOCK_SIZE_MIN_SECTORS ||
 	    block_size > DATA_DEV_BLOCK_SIZE_MAX_SECTORS ||
-	    !is_power_of_2(block_size)) {
+	    block_size & (DATA_DEV_BLOCK_SIZE_MIN_SECTORS - 1)) {
 		ti->error = "Invalid block size";
 		r = -EINVAL;
 		goto out;
@@ -1993,7 +2227,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 	}
 
 	pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev,
-			   block_size, &ti->error, &pool_created);
+			   block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created);
 	if (IS_ERR(pool)) {
 		r = PTR_ERR(pool);
 		goto out_free_pt;
@@ -2011,6 +2245,15 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		goto out_flags_changed;
 	}
 
+	/*
+	 * The block layer requires discard_granularity to be a power of 2.
+	 */
+	if (pf.discard_enabled && !is_power_of_2(block_size)) {
+		ti->error = "Discard support must be disabled when the block size is not a power of 2";
+		r = -EINVAL;
+		goto out_flags_changed;
+	}
+
 	pt->pool = pool;
 	pt->ti = ti;
 	pt->metadata_dev = metadata_dev;
@@ -2030,7 +2273,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		 * stacking of discard limits (this keeps the pool and
 		 * thin devices' discard limits consistent).
 		 */
-		ti->discards_supported = 1;
+		ti->discards_supported = true;
 	}
 	ti->private = pt;
 
@@ -2090,7 +2333,8 @@ static int pool_preresume(struct dm_target *ti)
 	int r;
 	struct pool_c *pt = ti->private;
 	struct pool *pool = pt->pool;
-	dm_block_t data_size, sb_data_size;
+	sector_t data_size = ti->len;
+	dm_block_t sb_data_size;
 
 	/*
 	 * Take control of the pool object.
@@ -2099,7 +2343,8 @@ static int pool_preresume(struct dm_target *ti)
 	if (r)
 		return r;
 
-	data_size = ti->len >> pool->block_shift;
+	(void) sector_div(data_size, pool->sectors_per_block);
+
 	r = dm_pool_get_data_dev_size(pool->pmd, &sb_data_size);
 	if (r) {
 		DMERR("failed to retrieve data device size");
@@ -2108,22 +2353,19 @@ static int pool_preresume(struct dm_target *ti)
 
 	if (data_size < sb_data_size) {
 		DMERR("pool target too small, is %llu blocks (expected %llu)",
-		      data_size, sb_data_size);
+		      (unsigned long long)data_size, sb_data_size);
 		return -EINVAL;
 
 	} else if (data_size > sb_data_size) {
 		r = dm_pool_resize_data_dev(pool->pmd, data_size);
 		if (r) {
 			DMERR("failed to resize data device");
+			/* FIXME Stricter than necessary: Rollback transaction instead here */
+			set_pool_mode(pool, PM_READ_ONLY);
 			return r;
 		}
 
-		r = dm_pool_commit_metadata(pool->pmd);
-		if (r) {
-			DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
-			      __func__, r);
-			return r;
-		}
+		(void) commit_or_fallback(pool);
 	}
 
 	return 0;
@@ -2146,19 +2388,12 @@ static void pool_resume(struct dm_target *ti)
 
 static void pool_postsuspend(struct dm_target *ti)
 {
-	int r;
 	struct pool_c *pt = ti->private;
 	struct pool *pool = pt->pool;
 
 	cancel_delayed_work(&pool->waker);
 	flush_workqueue(pool->wq);
-
-	r = dm_pool_commit_metadata(pool->pmd);
-	if (r < 0) {
-		DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
-		      __func__, r);
-		/* FIXME: invalidate device? error the next FUA or FLUSH bio ?*/
-	}
+	(void) commit_or_fallback(pool);
 }
 
 static int check_arg_count(unsigned argc, unsigned args_required)
@@ -2292,6 +2527,8 @@ static int process_reserve_metadata_snap_mesg(unsigned argc, char **argv, struct
 	if (r)
 		return r;
 
+	(void) commit_or_fallback(pool);
+
 	r = dm_pool_reserve_metadata_snap(pool->pmd);
 	if (r)
 		DMWARN("reserve_metadata_snap message failed.");
@@ -2351,25 +2588,41 @@ static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
 	else
 		DMWARN("Unrecognised thin pool target message received: %s", argv[0]);
 
-	if (!r) {
-		r = dm_pool_commit_metadata(pool->pmd);
-		if (r)
-			DMERR("%s message: dm_pool_commit_metadata() failed, error = %d",
-			      argv[0], r);
-	}
+	if (!r)
+		(void) commit_or_fallback(pool);
 
 	return r;
 }
 
+static void emit_flags(struct pool_features *pf, char *result,
+		       unsigned sz, unsigned maxlen)
+{
+	unsigned count = !pf->zero_new_blocks + !pf->discard_enabled +
+		!pf->discard_passdown + (pf->mode == PM_READ_ONLY);
+	DMEMIT("%u ", count);
+
+	if (!pf->zero_new_blocks)
+		DMEMIT("skip_block_zeroing ");
+
+	if (!pf->discard_enabled)
+		DMEMIT("ignore_discard ");
+
+	if (!pf->discard_passdown)
+		DMEMIT("no_discard_passdown ");
+
+	if (pf->mode == PM_READ_ONLY)
+		DMEMIT("read_only ");
+}
+
 /*
  * Status line is:
  *    <transaction id> <used metadata sectors>/<total metadata sectors>
  *    <used data sectors>/<total data sectors> <held metadata root>
  */
 static int pool_status(struct dm_target *ti, status_type_t type,
-		       char *result, unsigned maxlen)
+		       unsigned status_flags, char *result, unsigned maxlen)
 {
-	int r, count;
+	int r;
 	unsigned sz = 0;
 	uint64_t transaction_id;
 	dm_block_t nr_free_blocks_data;
@@ -2384,6 +2637,15 @@ static int pool_status(struct dm_target *ti, status_type_t type,
 
 	switch (type) {
 	case STATUSTYPE_INFO:
+		if (get_pool_mode(pool) == PM_FAIL) {
+			DMEMIT("Fail");
+			break;
+		}
+
+		/* Commit to ensure statistics aren't out-of-date */
+		if (!(status_flags & DM_STATUS_NOFLUSH_FLAG) && !dm_suspended(ti))
+			(void) commit_or_fallback(pool);
+
 		r = dm_pool_get_metadata_transaction_id(pool->pmd,
 							&transaction_id);
 		if (r)
@@ -2419,9 +2681,19 @@ static int pool_status(struct dm_target *ti, status_type_t type,
 		       (unsigned long long)nr_blocks_data);
 
 		if (held_root)
-			DMEMIT("%llu", held_root);
+			DMEMIT("%llu ", held_root);
+		else
+			DMEMIT("- ");
+
+		if (pool->pf.mode == PM_READ_ONLY)
+			DMEMIT("ro ");
+		else
+			DMEMIT("rw ");
+
+		if (pool->pf.discard_enabled && pool->pf.discard_passdown)
+			DMEMIT("discard_passdown");
 		else
-			DMEMIT("-");
+			DMEMIT("no_discard_passdown");
 
 		break;
 
@@ -2431,20 +2703,7 @@ static int pool_status(struct dm_target *ti, status_type_t type,
 		       format_dev_t(buf2, pt->data_dev->bdev->bd_dev),
 		       (unsigned long)pool->sectors_per_block,
 		       (unsigned long long)pt->low_water_blocks);
-
-		count = !pool->pf.zero_new_blocks + !pool->pf.discard_enabled +
-			!pt->pf.discard_passdown;
-		DMEMIT("%u ", count);
-
-		if (!pool->pf.zero_new_blocks)
-			DMEMIT("skip_block_zeroing ");
-
-		if (!pool->pf.discard_enabled)
-			DMEMIT("ignore_discard ");
-
-		if (!pt->pf.discard_passdown)
-			DMEMIT("no_discard_passdown ");
-
+		emit_flags(&pt->pf, result, sz, maxlen);
 		break;
 	}
 
@@ -2482,7 +2741,8 @@ static void set_discard_limits(struct pool *pool, struct queue_limits *limits)
 
 	/*
 	 * This is just a hint, and not enforced.  We have to cope with
-	 * bios that overlap 2 blocks.
+	 * bios that cover a block partially.  A discard that spans a block
+	 * boundary is not sent to this target.
 	 */
 	limits->discard_granularity = pool->sectors_per_block << SECTOR_SHIFT;
 	limits->discard_zeroes_data = pool->pf.zero_new_blocks;
@@ -2503,7 +2763,7 @@ static struct target_type pool_target = {
 	.name = "thin-pool",
 	.features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
 		    DM_TARGET_IMMUTABLE,
-	.version = {1, 2, 0},
+	.version = {1, 3, 0},
 	.module = THIS_MODULE,
 	.ctr = pool_ctr,
 	.dtr = pool_dtr,
@@ -2608,19 +2868,31 @@ static int thin_ctr(struct dm_target *ti, unsigned argc, char **argv)
 	}
 	__pool_inc(tc->pool);
 
+	if (get_pool_mode(tc->pool) == PM_FAIL) {
+		ti->error = "Couldn't open thin device, Pool is in fail mode";
+		goto bad_thin_open;
+	}
+
 	r = dm_pool_open_thin_device(tc->pool->pmd, tc->dev_id, &tc->td);
 	if (r) {
 		ti->error = "Couldn't open thin internal device";
 		goto bad_thin_open;
 	}
 
-	ti->split_io = tc->pool->sectors_per_block;
+	r = dm_set_target_max_io_len(ti, tc->pool->sectors_per_block);
+	if (r)
+		goto bad_thin_open;
+
 	ti->num_flush_requests = 1;
+	ti->flush_supported = true;
 
 	/* In case the pool supports discards, pass them on. */
 	if (tc->pool->pf.discard_enabled) {
-		ti->discards_supported = 1;
+		ti->discards_supported = true;
 		ti->num_discard_requests = 1;
+		ti->discard_zeroes_data_unsupported = true;
+		/* Discard requests must be split on a block boundary */
+		ti->split_discard_requests = true;
 	}
 
 	dm_put(pool_md);
@@ -2701,7 +2973,7 @@ static void thin_postsuspend(struct dm_target *ti)
  * <nr mapped sectors> <highest mapped sector>
  */
 static int thin_status(struct dm_target *ti, status_type_t type,
-		       char *result, unsigned maxlen)
+		       unsigned status_flags, char *result, unsigned maxlen)
 {
 	int r;
 	ssize_t sz = 0;
@@ -2709,6 +2981,11 @@ static int thin_status(struct dm_target *ti, status_type_t type,
 	char buf[BDEVNAME_SIZE];
 	struct thin_c *tc = ti->private;
 
+	if (get_pool_mode(tc->pool) == PM_FAIL) {
+		DMEMIT("Fail");
+		return 0;
+	}
+
 	if (!tc->td)
 		DMEMIT("-");
 	else {
@@ -2746,19 +3023,21 @@ static int thin_status(struct dm_target *ti, status_type_t type,
 static int thin_iterate_devices(struct dm_target *ti,
 				iterate_devices_callout_fn fn, void *data)
 {
-	dm_block_t blocks;
+	sector_t blocks;
 	struct thin_c *tc = ti->private;
+	struct pool *pool = tc->pool;
 
 	/*
 	 * We can't call dm_pool_get_data_dev_size() since that blocks.  So
 	 * we follow a more convoluted path through to the pool's target.
 	 */
-	if (!tc->pool->ti)
+	if (!pool->ti)
 		return 0;	/* nothing is bound */
 
-	blocks = tc->pool->ti->len >> tc->pool->block_shift;
+	blocks = pool->ti->len;
+	(void) sector_div(blocks, pool->sectors_per_block);
 	if (blocks)
-		return fn(ti, tc->pool_dev, 0, tc->pool->sectors_per_block * blocks, data);
+		return fn(ti, tc->pool_dev, 0, pool->sectors_per_block * blocks, data);
 
 	return 0;
 }
@@ -2775,7 +3054,7 @@ static void thin_io_hints(struct dm_target *ti, struct queue_limits *limits)
 
 static struct target_type thin_target = {
 	.name = "thin",
-	.version = {1, 1, 0},
+	.version = {1, 3, 0},
 	.module	= THIS_MODULE,
 	.ctr = thin_ctr,
 	.dtr = thin_dtr,
diff --git a/drivers/md/dm-verity.c b/drivers/md/dm-verity.c
index fa365d39b61..254d19268ad 100644
--- a/drivers/md/dm-verity.c
+++ b/drivers/md/dm-verity.c
@@ -515,7 +515,7 @@ static int verity_map(struct dm_target *ti, struct bio *bio,
  * Status: V (valid) or C (corruption found)
  */
 static int verity_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned maxlen)
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct dm_verity *v = ti->private;
 	unsigned sz = 0;
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index e24143cc204..4e09b6ff5b4 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -968,22 +968,41 @@ static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti
 static sector_t max_io_len(sector_t sector, struct dm_target *ti)
 {
 	sector_t len = max_io_len_target_boundary(sector, ti);
+	sector_t offset, max_len;
 
 	/*
-	 * Does the target need to split even further ?
+	 * Does the target need to split even further?
 	 */
-	if (ti->split_io) {
-		sector_t boundary;
-		sector_t offset = dm_target_offset(ti, sector);
-		boundary = ((offset + ti->split_io) & ~(ti->split_io - 1))
-			   - offset;
-		if (len > boundary)
-			len = boundary;
+	if (ti->max_io_len) {
+		offset = dm_target_offset(ti, sector);
+		if (unlikely(ti->max_io_len & (ti->max_io_len - 1)))
+			max_len = sector_div(offset, ti->max_io_len);
+		else
+			max_len = offset & (ti->max_io_len - 1);
+		max_len = ti->max_io_len - max_len;
+
+		if (len > max_len)
+			len = max_len;
 	}
 
 	return len;
 }
 
+int dm_set_target_max_io_len(struct dm_target *ti, sector_t len)
+{
+	if (len > UINT_MAX) {
+		DMERR("Specified maximum size of target IO (%llu) exceeds limit (%u)",
+		      (unsigned long long)len, UINT_MAX);
+		ti->error = "Maximum size of target IO is too large";
+		return -EINVAL;
+	}
+
+	ti->max_io_len = (uint32_t) len;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_set_target_max_io_len);
+
 static void __map_bio(struct dm_target *ti, struct bio *clone,
 		      struct dm_target_io *tio)
 {
@@ -1196,7 +1215,10 @@ static int __clone_and_map_discard(struct clone_info *ci)
 		if (!ti->num_discard_requests)
 			return -EOPNOTSUPP;
 
-		len = min(ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
+		if (!ti->split_discard_requests)
+			len = min(ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
+		else
+			len = min(ci->sector_count, max_io_len(ci->sector, ti));
 
 		__issue_target_requests(ci, ti, ti->num_discard_requests, len);
 
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index b7dacd59d8d..52eef493d26 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -23,6 +23,11 @@
 #define DM_SUSPEND_NOFLUSH_FLAG		(1 << 1)
 
 /*
+ * Status feature flags
+ */
+#define DM_STATUS_NOFLUSH_FLAG		(1 << 0)
+
+/*
  * Type of table and mapped_device's mempool
  */
 #define DM_TYPE_NONE		0
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 1c2f9048e1a..fcd098794d3 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -498,61 +498,13 @@ void md_flush_request(struct mddev *mddev, struct bio *bio)
 }
 EXPORT_SYMBOL(md_flush_request);
 
-/* Support for plugging.
- * This mirrors the plugging support in request_queue, but does not
- * require having a whole queue or request structures.
- * We allocate an md_plug_cb for each md device and each thread it gets
- * plugged on.  This links tot the private plug_handle structure in the
- * personality data where we keep a count of the number of outstanding
- * plugs so other code can see if a plug is active.
- */
-struct md_plug_cb {
-	struct blk_plug_cb cb;
-	struct mddev *mddev;
-};
-
-static void plugger_unplug(struct blk_plug_cb *cb)
-{
-	struct md_plug_cb *mdcb = container_of(cb, struct md_plug_cb, cb);
-	if (atomic_dec_and_test(&mdcb->mddev->plug_cnt))
-		md_wakeup_thread(mdcb->mddev->thread);
-	kfree(mdcb);
-}
-
-/* Check that an unplug wakeup will come shortly.
- * If not, wakeup the md thread immediately
- */
-int mddev_check_plugged(struct mddev *mddev)
+void md_unplug(struct blk_plug_cb *cb, bool from_schedule)
 {
-	struct blk_plug *plug = current->plug;
-	struct md_plug_cb *mdcb;
-
-	if (!plug)
-		return 0;
-
-	list_for_each_entry(mdcb, &plug->cb_list, cb.list) {
-		if (mdcb->cb.callback == plugger_unplug &&
-		    mdcb->mddev == mddev) {
-			/* Already on the list, move to top */
-			if (mdcb != list_first_entry(&plug->cb_list,
-						    struct md_plug_cb,
-						    cb.list))
-				list_move(&mdcb->cb.list, &plug->cb_list);
-			return 1;
-		}
-	}
-	/* Not currently on the callback list */
-	mdcb = kmalloc(sizeof(*mdcb), GFP_ATOMIC);
-	if (!mdcb)
-		return 0;
-
-	mdcb->mddev = mddev;
-	mdcb->cb.callback = plugger_unplug;
-	atomic_inc(&mddev->plug_cnt);
-	list_add(&mdcb->cb.list, &plug->cb_list);
-	return 1;
+	struct mddev *mddev = cb->data;
+	md_wakeup_thread(mddev->thread);
+	kfree(cb);
 }
-EXPORT_SYMBOL_GPL(mddev_check_plugged);
+EXPORT_SYMBOL(md_unplug);
 
 static inline struct mddev *mddev_get(struct mddev *mddev)
 {
@@ -602,7 +554,6 @@ void mddev_init(struct mddev *mddev)
 	atomic_set(&mddev->active, 1);
 	atomic_set(&mddev->openers, 0);
 	atomic_set(&mddev->active_io, 0);
-	atomic_set(&mddev->plug_cnt, 0);
 	spin_lock_init(&mddev->write_lock);
 	atomic_set(&mddev->flush_pending, 0);
 	init_waitqueue_head(&mddev->sb_wait);
@@ -2931,6 +2882,7 @@ offset_store(struct md_rdev *rdev, const char *buf, size_t len)
 		 * can be sane */
 		return -EBUSY;
 	rdev->data_offset = offset;
+	rdev->new_data_offset = offset;
 	return len;
 }
 
@@ -3926,8 +3878,8 @@ array_state_show(struct mddev *mddev, char *page)
 	return sprintf(page, "%s\n", array_states[st]);
 }
 
-static int do_md_stop(struct mddev * mddev, int ro, int is_open);
-static int md_set_readonly(struct mddev * mddev, int is_open);
+static int do_md_stop(struct mddev * mddev, int ro, struct block_device *bdev);
+static int md_set_readonly(struct mddev * mddev, struct block_device *bdev);
 static int do_md_run(struct mddev * mddev);
 static int restart_array(struct mddev *mddev);
 
@@ -3941,24 +3893,20 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 		break;
 	case clear:
 		/* stopping an active array */
-		if (atomic_read(&mddev->openers) > 0)
-			return -EBUSY;
-		err = do_md_stop(mddev, 0, 0);
+		err = do_md_stop(mddev, 0, NULL);
 		break;
 	case inactive:
 		/* stopping an active array */
-		if (mddev->pers) {
-			if (atomic_read(&mddev->openers) > 0)
-				return -EBUSY;
-			err = do_md_stop(mddev, 2, 0);
-		} else
+		if (mddev->pers)
+			err = do_md_stop(mddev, 2, NULL);
+		else
 			err = 0; /* already inactive */
 		break;
 	case suspended:
 		break; /* not supported yet */
 	case readonly:
 		if (mddev->pers)
-			err = md_set_readonly(mddev, 0);
+			err = md_set_readonly(mddev, NULL);
 		else {
 			mddev->ro = 1;
 			set_disk_ro(mddev->gendisk, 1);
@@ -3968,7 +3916,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 	case read_auto:
 		if (mddev->pers) {
 			if (mddev->ro == 0)
-				err = md_set_readonly(mddev, 0);
+				err = md_set_readonly(mddev, NULL);
 			else if (mddev->ro == 1)
 				err = restart_array(mddev);
 			if (err == 0) {
@@ -5351,15 +5299,17 @@ void md_stop(struct mddev *mddev)
 }
 EXPORT_SYMBOL_GPL(md_stop);
 
-static int md_set_readonly(struct mddev *mddev, int is_open)
+static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
 {
 	int err = 0;
 	mutex_lock(&mddev->open_mutex);
-	if (atomic_read(&mddev->openers) > is_open) {
+	if (atomic_read(&mddev->openers) > !!bdev) {
 		printk("md: %s still in use.\n",mdname(mddev));
 		err = -EBUSY;
 		goto out;
 	}
+	if (bdev)
+		sync_blockdev(bdev);
 	if (mddev->pers) {
 		__md_stop_writes(mddev);
 
@@ -5381,18 +5331,26 @@ out:
  *   0 - completely stop and dis-assemble array
  *   2 - stop but do not disassemble array
  */
-static int do_md_stop(struct mddev * mddev, int mode, int is_open)
+static int do_md_stop(struct mddev * mddev, int mode,
+		      struct block_device *bdev)
 {
 	struct gendisk *disk = mddev->gendisk;
 	struct md_rdev *rdev;
 
 	mutex_lock(&mddev->open_mutex);
-	if (atomic_read(&mddev->openers) > is_open ||
+	if (atomic_read(&mddev->openers) > !!bdev ||
 	    mddev->sysfs_active) {
 		printk("md: %s still in use.\n",mdname(mddev));
 		mutex_unlock(&mddev->open_mutex);
 		return -EBUSY;
 	}
+	if (bdev)
+		/* It is possible IO was issued on some other
+		 * open file which was closed before we took ->open_mutex.
+		 * As that was not the last close __blkdev_put will not
+		 * have called sync_blockdev, so we must.
+		 */
+		sync_blockdev(bdev);
 
 	if (mddev->pers) {
 		if (mddev->ro)
@@ -5466,7 +5424,7 @@ static void autorun_array(struct mddev *mddev)
 	err = do_md_run(mddev);
 	if (err) {
 		printk(KERN_WARNING "md: do_md_run() returned %d\n", err);
-		do_md_stop(mddev, 0, 0);
+		do_md_stop(mddev, 0, NULL);
 	}
 }
 
@@ -5784,8 +5742,7 @@ static int add_new_disk(struct mddev * mddev, mdu_disk_info_t *info)
 			super_types[mddev->major_version].
 				validate_super(mddev, rdev);
 		if ((info->state & (1<<MD_DISK_SYNC)) &&
-		    (!test_bit(In_sync, &rdev->flags) ||
-		     rdev->raid_disk != info->raid_disk)) {
+		     rdev->raid_disk != info->raid_disk) {
 			/* This was a hot-add request, but events doesn't
 			 * match, so reject it.
 			 */
@@ -6482,11 +6439,11 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 			goto done_unlock;
 
 		case STOP_ARRAY:
-			err = do_md_stop(mddev, 0, 1);
+			err = do_md_stop(mddev, 0, bdev);
 			goto done_unlock;
 
 		case STOP_ARRAY_RO:
-			err = md_set_readonly(mddev, 1);
+			err = md_set_readonly(mddev, bdev);
 			goto done_unlock;
 
 		case BLKROSET:
@@ -6751,7 +6708,7 @@ struct md_thread *md_register_thread(void (*run) (struct mddev *), struct mddev
 	thread->tsk = kthread_run(md_thread, thread,
 				  "%s_%s",
 				  mdname(thread->mddev),
-				  name ?: mddev->pers->name);
+				  name);
 	if (IS_ERR(thread->tsk)) {
 		kfree(thread);
 		return NULL;
@@ -7298,6 +7255,7 @@ void md_do_sync(struct mddev *mddev)
 	int skipped = 0;
 	struct md_rdev *rdev;
 	char *desc;
+	struct blk_plug plug;
 
 	/* just incase thread restarts... */
 	if (test_bit(MD_RECOVERY_DONE, &mddev->recovery))
@@ -7447,6 +7405,7 @@ void md_do_sync(struct mddev *mddev)
 	}
 	mddev->curr_resync_completed = j;
 
+	blk_start_plug(&plug);
 	while (j < max_sectors) {
 		sector_t sectors;
 
@@ -7552,6 +7511,7 @@ void md_do_sync(struct mddev *mddev)
 	 * this also signals 'finished resyncing' to md_stop
 	 */
  out:
+	blk_finish_plug(&plug);
 	wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
 
 	/* tell personality that we are finished */
diff --git a/drivers/md/md.h b/drivers/md/md.h
index 7b4a3c318ca..f385b038589 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -266,9 +266,6 @@ struct mddev {
 	int				new_chunk_sectors;
 	int				reshape_backwards;
 
-	atomic_t			plug_cnt;	/* If device is expecting
-							 * more bios soon.
-							 */
 	struct md_thread		*thread;	/* management thread */
 	struct md_thread		*sync_thread;	/* doing resync or reconstruct */
 	sector_t			curr_resync;	/* last block scheduled */
@@ -630,6 +627,12 @@ extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
 				   struct mddev *mddev);
 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 				   struct mddev *mddev);
-extern int mddev_check_plugged(struct mddev *mddev);
 extern void md_trim_bio(struct bio *bio, int offset, int size);
+
+extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
+static inline int mddev_check_plugged(struct mddev *mddev)
+{
+	return !!blk_check_plugged(md_unplug, mddev,
+				   sizeof(struct blk_plug_cb));
+}
 #endif /* _MD_MD_H */
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index 9339e67fcc7..61a1833ebaf 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -474,7 +474,8 @@ static int multipath_run (struct mddev *mddev)
 	}
 
 	{
-		mddev->thread = md_register_thread(multipathd, mddev, NULL);
+		mddev->thread = md_register_thread(multipathd, mddev,
+						   "multipath");
 		if (!mddev->thread) {
 			printk(KERN_ERR "multipath: couldn't allocate thread"
 				" for %s\n", mdname(mddev));
diff --git a/drivers/md/persistent-data/Makefile b/drivers/md/persistent-data/Makefile
index cfa95f66223..d8e7cb767c1 100644
--- a/drivers/md/persistent-data/Makefile
+++ b/drivers/md/persistent-data/Makefile
@@ -1,7 +1,6 @@
 obj-$(CONFIG_DM_PERSISTENT_DATA) += dm-persistent-data.o
 dm-persistent-data-objs := \
 	dm-block-manager.o \
-	dm-space-map-checker.o \
 	dm-space-map-common.o \
 	dm-space-map-disk.o \
 	dm-space-map-metadata.o \
diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c
index 0317ecdc6e5..5ba277768d9 100644
--- a/drivers/md/persistent-data/dm-block-manager.c
+++ b/drivers/md/persistent-data/dm-block-manager.c
@@ -325,11 +325,6 @@ static struct dm_buffer *to_buffer(struct dm_block *b)
 	return (struct dm_buffer *) b;
 }
 
-static struct dm_bufio_client *to_bufio(struct dm_block_manager *bm)
-{
-	return (struct dm_bufio_client *) bm;
-}
-
 dm_block_t dm_block_location(struct dm_block *b)
 {
 	return dm_bufio_get_block_number(to_buffer(b));
@@ -367,34 +362,60 @@ static void dm_block_manager_write_callback(struct dm_buffer *buf)
 /*----------------------------------------------------------------
  * Public interface
  *--------------------------------------------------------------*/
+struct dm_block_manager {
+	struct dm_bufio_client *bufio;
+	bool read_only:1;
+};
+
 struct dm_block_manager *dm_block_manager_create(struct block_device *bdev,
 						 unsigned block_size,
 						 unsigned cache_size,
 						 unsigned max_held_per_thread)
 {
-	return (struct dm_block_manager *)
-		dm_bufio_client_create(bdev, block_size, max_held_per_thread,
-				       sizeof(struct buffer_aux),
-				       dm_block_manager_alloc_callback,
-				       dm_block_manager_write_callback);
+	int r;
+	struct dm_block_manager *bm;
+
+	bm = kmalloc(sizeof(*bm), GFP_KERNEL);
+	if (!bm) {
+		r = -ENOMEM;
+		goto bad;
+	}
+
+	bm->bufio = dm_bufio_client_create(bdev, block_size, max_held_per_thread,
+					   sizeof(struct buffer_aux),
+					   dm_block_manager_alloc_callback,
+					   dm_block_manager_write_callback);
+	if (IS_ERR(bm->bufio)) {
+		r = PTR_ERR(bm->bufio);
+		kfree(bm);
+		goto bad;
+	}
+
+	bm->read_only = false;
+
+	return bm;
+
+bad:
+	return ERR_PTR(r);
 }
 EXPORT_SYMBOL_GPL(dm_block_manager_create);
 
 void dm_block_manager_destroy(struct dm_block_manager *bm)
 {
-	return dm_bufio_client_destroy(to_bufio(bm));
+	dm_bufio_client_destroy(bm->bufio);
+	kfree(bm);
 }
 EXPORT_SYMBOL_GPL(dm_block_manager_destroy);
 
 unsigned dm_bm_block_size(struct dm_block_manager *bm)
 {
-	return dm_bufio_get_block_size(to_bufio(bm));
+	return dm_bufio_get_block_size(bm->bufio);
 }
 EXPORT_SYMBOL_GPL(dm_bm_block_size);
 
 dm_block_t dm_bm_nr_blocks(struct dm_block_manager *bm)
 {
-	return dm_bufio_get_device_size(to_bufio(bm));
+	return dm_bufio_get_device_size(bm->bufio);
 }
 
 static int dm_bm_validate_buffer(struct dm_block_manager *bm,
@@ -406,7 +427,7 @@ static int dm_bm_validate_buffer(struct dm_block_manager *bm,
 		int r;
 		if (!v)
 			return 0;
-		r = v->check(v, (struct dm_block *) buf, dm_bufio_get_block_size(to_bufio(bm)));
+		r = v->check(v, (struct dm_block *) buf, dm_bufio_get_block_size(bm->bufio));
 		if (unlikely(r))
 			return r;
 		aux->validator = v;
@@ -430,7 +451,7 @@ int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
 	void *p;
 	int r;
 
-	p = dm_bufio_read(to_bufio(bm), b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
 	if (unlikely(IS_ERR(p)))
 		return PTR_ERR(p);
 
@@ -463,7 +484,10 @@ int dm_bm_write_lock(struct dm_block_manager *bm,
 	void *p;
 	int r;
 
-	p = dm_bufio_read(to_bufio(bm), b, (struct dm_buffer **) result);
+	if (bm->read_only)
+		return -EPERM;
+
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
 	if (unlikely(IS_ERR(p)))
 		return PTR_ERR(p);
 
@@ -496,7 +520,7 @@ int dm_bm_read_try_lock(struct dm_block_manager *bm,
 	void *p;
 	int r;
 
-	p = dm_bufio_get(to_bufio(bm), b, (struct dm_buffer **) result);
+	p = dm_bufio_get(bm->bufio, b, (struct dm_buffer **) result);
 	if (unlikely(IS_ERR(p)))
 		return PTR_ERR(p);
 	if (unlikely(!p))
@@ -529,7 +553,10 @@ int dm_bm_write_lock_zero(struct dm_block_manager *bm,
 	struct buffer_aux *aux;
 	void *p;
 
-	p = dm_bufio_new(to_bufio(bm), b, (struct dm_buffer **) result);
+	if (bm->read_only)
+		return -EPERM;
+
+	p = dm_bufio_new(bm->bufio, b, (struct dm_buffer **) result);
 	if (unlikely(IS_ERR(p)))
 		return PTR_ERR(p);
 
@@ -547,6 +574,7 @@ int dm_bm_write_lock_zero(struct dm_block_manager *bm,
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(dm_bm_write_lock_zero);
 
 int dm_bm_unlock(struct dm_block *b)
 {
@@ -565,45 +593,30 @@ int dm_bm_unlock(struct dm_block *b)
 }
 EXPORT_SYMBOL_GPL(dm_bm_unlock);
 
-int dm_bm_unlock_move(struct dm_block *b, dm_block_t n)
-{
-	struct buffer_aux *aux;
-
-	aux = dm_bufio_get_aux_data(to_buffer(b));
-
-	if (aux->write_locked) {
-		dm_bufio_mark_buffer_dirty(to_buffer(b));
-		bl_up_write(&aux->lock);
-	} else
-		bl_up_read(&aux->lock);
-
-	dm_bufio_release_move(to_buffer(b), n);
-	return 0;
-}
-
 int dm_bm_flush_and_unlock(struct dm_block_manager *bm,
 			   struct dm_block *superblock)
 {
 	int r;
 
-	r = dm_bufio_write_dirty_buffers(to_bufio(bm));
-	if (unlikely(r))
-		return r;
-	r = dm_bufio_issue_flush(to_bufio(bm));
-	if (unlikely(r))
+	if (bm->read_only)
+		return -EPERM;
+
+	r = dm_bufio_write_dirty_buffers(bm->bufio);
+	if (unlikely(r)) {
+		dm_bm_unlock(superblock);
 		return r;
+	}
 
 	dm_bm_unlock(superblock);
 
-	r = dm_bufio_write_dirty_buffers(to_bufio(bm));
-	if (unlikely(r))
-		return r;
-	r = dm_bufio_issue_flush(to_bufio(bm));
-	if (unlikely(r))
-		return r;
+	return dm_bufio_write_dirty_buffers(bm->bufio);
+}
 
-	return 0;
+void dm_bm_set_read_only(struct dm_block_manager *bm)
+{
+	bm->read_only = true;
 }
+EXPORT_SYMBOL_GPL(dm_bm_set_read_only);
 
 u32 dm_bm_checksum(const void *data, size_t len, u32 init_xor)
 {
diff --git a/drivers/md/persistent-data/dm-block-manager.h b/drivers/md/persistent-data/dm-block-manager.h
index 924833d2dfa..be5bff61be2 100644
--- a/drivers/md/persistent-data/dm-block-manager.h
+++ b/drivers/md/persistent-data/dm-block-manager.h
@@ -97,14 +97,6 @@ int dm_bm_write_lock_zero(struct dm_block_manager *bm, dm_block_t b,
 int dm_bm_unlock(struct dm_block *b);
 
 /*
- * An optimisation; we often want to copy a block's contents to a new
- * block.  eg, as part of the shadowing operation.  It's far better for
- * bufio to do this move behind the scenes than hold 2 locks and memcpy the
- * data.
- */
-int dm_bm_unlock_move(struct dm_block *b, dm_block_t n);
-
-/*
  * It's a common idiom to have a superblock that should be committed last.
  *
  * @superblock should be write-locked on entry. It will be unlocked during
@@ -116,6 +108,19 @@ int dm_bm_unlock_move(struct dm_block *b, dm_block_t n);
 int dm_bm_flush_and_unlock(struct dm_block_manager *bm,
 			   struct dm_block *superblock);
 
+/*
+ * Switches the bm to a read only mode.  Once read-only mode
+ * has been entered the following functions will return -EPERM.
+ *
+ *   dm_bm_write_lock
+ *   dm_bm_write_lock_zero
+ *   dm_bm_flush_and_unlock
+ *
+ * Additionally you should not use dm_bm_unlock_move, however no error will
+ * be returned if you do.
+ */
+void dm_bm_set_read_only(struct dm_block_manager *bm);
+
 u32 dm_bm_checksum(const void *data, size_t len, u32 init_xor);
 
 /*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-space-map-checker.c b/drivers/md/persistent-data/dm-space-map-checker.c
deleted file mode 100644
index 50ed53bf4aa..00000000000
--- a/drivers/md/persistent-data/dm-space-map-checker.c
+++ /dev/null
@@ -1,438 +0,0 @@
-/*
- * Copyright (C) 2011 Red Hat, Inc.
- *
- * This file is released under the GPL.
- */
-
-#include "dm-space-map-checker.h"
-
-#include <linux/device-mapper.h>
-#include <linux/export.h>
-
-#ifdef CONFIG_DM_DEBUG_SPACE_MAPS
-
-#define DM_MSG_PREFIX "space map checker"
-
-/*----------------------------------------------------------------*/
-
-struct count_array {
-	dm_block_t nr;
-	dm_block_t nr_free;
-
-	uint32_t *counts;
-};
-
-static int ca_get_count(struct count_array *ca, dm_block_t b, uint32_t *count)
-{
-	if (b >= ca->nr)
-		return -EINVAL;
-
-	*count = ca->counts[b];
-	return 0;
-}
-
-static int ca_count_more_than_one(struct count_array *ca, dm_block_t b, int *r)
-{
-	if (b >= ca->nr)
-		return -EINVAL;
-
-	*r = ca->counts[b] > 1;
-	return 0;
-}
-
-static int ca_set_count(struct count_array *ca, dm_block_t b, uint32_t count)
-{
-	uint32_t old_count;
-
-	if (b >= ca->nr)
-		return -EINVAL;
-
-	old_count = ca->counts[b];
-
-	if (!count && old_count)
-		ca->nr_free++;
-
-	else if (count && !old_count)
-		ca->nr_free--;
-
-	ca->counts[b] = count;
-	return 0;
-}
-
-static int ca_inc_block(struct count_array *ca, dm_block_t b)
-{
-	if (b >= ca->nr)
-		return -EINVAL;
-
-	ca_set_count(ca, b, ca->counts[b] + 1);
-	return 0;
-}
-
-static int ca_dec_block(struct count_array *ca, dm_block_t b)
-{
-	if (b >= ca->nr)
-		return -EINVAL;
-
-	BUG_ON(ca->counts[b] == 0);
-	ca_set_count(ca, b, ca->counts[b] - 1);
-	return 0;
-}
-
-static int ca_create(struct count_array *ca, struct dm_space_map *sm)
-{
-	int r;
-	dm_block_t nr_blocks;
-
-	r = dm_sm_get_nr_blocks(sm, &nr_blocks);
-	if (r)
-		return r;
-
-	ca->nr = nr_blocks;
-	ca->nr_free = nr_blocks;
-	ca->counts = kzalloc(sizeof(*ca->counts) * nr_blocks, GFP_KERNEL);
-	if (!ca->counts)
-		return -ENOMEM;
-
-	return 0;
-}
-
-static int ca_load(struct count_array *ca, struct dm_space_map *sm)
-{
-	int r;
-	uint32_t count;
-	dm_block_t nr_blocks, i;
-
-	r = dm_sm_get_nr_blocks(sm, &nr_blocks);
-	if (r)
-		return r;
-
-	BUG_ON(ca->nr != nr_blocks);
-
-	DMWARN("Loading debug space map from disk.  This may take some time");
-	for (i = 0; i < nr_blocks; i++) {
-		r = dm_sm_get_count(sm, i, &count);
-		if (r) {
-			DMERR("load failed");
-			return r;
-		}
-
-		ca_set_count(ca, i, count);
-	}
-	DMWARN("Load complete");
-
-	return 0;
-}
-
-static int ca_extend(struct count_array *ca, dm_block_t extra_blocks)
-{
-	dm_block_t nr_blocks = ca->nr + extra_blocks;
-	uint32_t *counts = kzalloc(sizeof(*counts) * nr_blocks, GFP_KERNEL);
-	if (!counts)
-		return -ENOMEM;
-
-	memcpy(counts, ca->counts, sizeof(*counts) * ca->nr);
-	kfree(ca->counts);
-	ca->nr = nr_blocks;
-	ca->nr_free += extra_blocks;
-	ca->counts = counts;
-	return 0;
-}
-
-static int ca_commit(struct count_array *old, struct count_array *new)
-{
-	if (old->nr != new->nr) {
-		BUG_ON(old->nr > new->nr);
-		ca_extend(old, new->nr - old->nr);
-	}
-
-	BUG_ON(old->nr != new->nr);
-	old->nr_free = new->nr_free;
-	memcpy(old->counts, new->counts, sizeof(*old->counts) * old->nr);
-	return 0;
-}
-
-static void ca_destroy(struct count_array *ca)
-{
-	kfree(ca->counts);
-}
-
-/*----------------------------------------------------------------*/
-
-struct sm_checker {
-	struct dm_space_map sm;
-
-	struct count_array old_counts;
-	struct count_array counts;
-
-	struct dm_space_map *real_sm;
-};
-
-static void sm_checker_destroy(struct dm_space_map *sm)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-
-	dm_sm_destroy(smc->real_sm);
-	ca_destroy(&smc->old_counts);
-	ca_destroy(&smc->counts);
-	kfree(smc);
-}
-
-static int sm_checker_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_get_nr_blocks(smc->real_sm, count);
-	if (!r)
-		BUG_ON(smc->old_counts.nr != *count);
-	return r;
-}
-
-static int sm_checker_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_get_nr_free(smc->real_sm, count);
-	if (!r) {
-		/*
-		 * Slow, but we know it's correct.
-		 */
-		dm_block_t b, n = 0;
-		for (b = 0; b < smc->old_counts.nr; b++)
-			if (smc->old_counts.counts[b] == 0 &&
-			    smc->counts.counts[b] == 0)
-				n++;
-
-		if (n != *count)
-			DMERR("free block counts differ, checker %u, sm-disk:%u",
-			      (unsigned) n, (unsigned) *count);
-	}
-	return r;
-}
-
-static int sm_checker_new_block(struct dm_space_map *sm, dm_block_t *b)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_new_block(smc->real_sm, b);
-
-	if (!r) {
-		BUG_ON(*b >= smc->old_counts.nr);
-		BUG_ON(smc->old_counts.counts[*b] != 0);
-		BUG_ON(*b >= smc->counts.nr);
-		BUG_ON(smc->counts.counts[*b] != 0);
-		ca_set_count(&smc->counts, *b, 1);
-	}
-
-	return r;
-}
-
-static int sm_checker_inc_block(struct dm_space_map *sm, dm_block_t b)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_inc_block(smc->real_sm, b);
-	int r2 = ca_inc_block(&smc->counts, b);
-	BUG_ON(r != r2);
-	return r;
-}
-
-static int sm_checker_dec_block(struct dm_space_map *sm, dm_block_t b)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_dec_block(smc->real_sm, b);
-	int r2 = ca_dec_block(&smc->counts, b);
-	BUG_ON(r != r2);
-	return r;
-}
-
-static int sm_checker_get_count(struct dm_space_map *sm, dm_block_t b, uint32_t *result)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	uint32_t result2 = 0;
-	int r = dm_sm_get_count(smc->real_sm, b, result);
-	int r2 = ca_get_count(&smc->counts, b, &result2);
-
-	BUG_ON(r != r2);
-	if (!r)
-		BUG_ON(*result != result2);
-	return r;
-}
-
-static int sm_checker_count_more_than_one(struct dm_space_map *sm, dm_block_t b, int *result)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int result2 = 0;
-	int r = dm_sm_count_is_more_than_one(smc->real_sm, b, result);
-	int r2 = ca_count_more_than_one(&smc->counts, b, &result2);
-
-	BUG_ON(r != r2);
-	if (!r)
-		BUG_ON(!(*result) && result2);
-	return r;
-}
-
-static int sm_checker_set_count(struct dm_space_map *sm, dm_block_t b, uint32_t count)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	uint32_t old_rc;
-	int r = dm_sm_set_count(smc->real_sm, b, count);
-	int r2;
-
-	BUG_ON(b >= smc->counts.nr);
-	old_rc = smc->counts.counts[b];
-	r2 = ca_set_count(&smc->counts, b, count);
-	BUG_ON(r != r2);
-
-	return r;
-}
-
-static int sm_checker_commit(struct dm_space_map *sm)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r;
-
-	r = dm_sm_commit(smc->real_sm);
-	if (r)
-		return r;
-
-	r = ca_commit(&smc->old_counts, &smc->counts);
-	if (r)
-		return r;
-
-	return 0;
-}
-
-static int sm_checker_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	int r = dm_sm_extend(smc->real_sm, extra_blocks);
-	if (r)
-		return r;
-
-	return ca_extend(&smc->counts, extra_blocks);
-}
-
-static int sm_checker_root_size(struct dm_space_map *sm, size_t *result)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	return dm_sm_root_size(smc->real_sm, result);
-}
-
-static int sm_checker_copy_root(struct dm_space_map *sm, void *copy_to_here_le, size_t len)
-{
-	struct sm_checker *smc = container_of(sm, struct sm_checker, sm);
-	return dm_sm_copy_root(smc->real_sm, copy_to_here_le, len);
-}
-
-/*----------------------------------------------------------------*/
-
-static struct dm_space_map ops_ = {
-	.destroy = sm_checker_destroy,
-	.get_nr_blocks = sm_checker_get_nr_blocks,
-	.get_nr_free = sm_checker_get_nr_free,
-	.inc_block = sm_checker_inc_block,
-	.dec_block = sm_checker_dec_block,
-	.new_block = sm_checker_new_block,
-	.get_count = sm_checker_get_count,
-	.count_is_more_than_one = sm_checker_count_more_than_one,
-	.set_count = sm_checker_set_count,
-	.commit = sm_checker_commit,
-	.extend = sm_checker_extend,
-	.root_size = sm_checker_root_size,
-	.copy_root = sm_checker_copy_root
-};
-
-struct dm_space_map *dm_sm_checker_create(struct dm_space_map *sm)
-{
-	int r;
-	struct sm_checker *smc;
-
-	if (!sm)
-		return NULL;
-
-	smc = kmalloc(sizeof(*smc), GFP_KERNEL);
-	if (!smc)
-		return NULL;
-
-	memcpy(&smc->sm, &ops_, sizeof(smc->sm));
-	r = ca_create(&smc->old_counts, sm);
-	if (r) {
-		kfree(smc);
-		return NULL;
-	}
-
-	r = ca_create(&smc->counts, sm);
-	if (r) {
-		ca_destroy(&smc->old_counts);
-		kfree(smc);
-		return NULL;
-	}
-
-	smc->real_sm = sm;
-
-	r = ca_load(&smc->counts, sm);
-	if (r) {
-		ca_destroy(&smc->counts);
-		ca_destroy(&smc->old_counts);
-		kfree(smc);
-		return NULL;
-	}
-
-	r = ca_commit(&smc->old_counts, &smc->counts);
-	if (r) {
-		ca_destroy(&smc->counts);
-		ca_destroy(&smc->old_counts);
-		kfree(smc);
-		return NULL;
-	}
-
-	return &smc->sm;
-}
-EXPORT_SYMBOL_GPL(dm_sm_checker_create);
-
-struct dm_space_map *dm_sm_checker_create_fresh(struct dm_space_map *sm)
-{
-	int r;
-	struct sm_checker *smc;
-
-	if (!sm)
-		return NULL;
-
-	smc = kmalloc(sizeof(*smc), GFP_KERNEL);
-	if (!smc)
-		return NULL;
-
-	memcpy(&smc->sm, &ops_, sizeof(smc->sm));
-	r = ca_create(&smc->old_counts, sm);
-	if (r) {
-		kfree(smc);
-		return NULL;
-	}
-
-	r = ca_create(&smc->counts, sm);
-	if (r) {
-		ca_destroy(&smc->old_counts);
-		kfree(smc);
-		return NULL;
-	}
-
-	smc->real_sm = sm;
-	return &smc->sm;
-}
-EXPORT_SYMBOL_GPL(dm_sm_checker_create_fresh);
-
-/*----------------------------------------------------------------*/
-
-#else
-
-struct dm_space_map *dm_sm_checker_create(struct dm_space_map *sm)
-{
-	return sm;
-}
-EXPORT_SYMBOL_GPL(dm_sm_checker_create);
-
-struct dm_space_map *dm_sm_checker_create_fresh(struct dm_space_map *sm)
-{
-	return sm;
-}
-EXPORT_SYMBOL_GPL(dm_sm_checker_create_fresh);
-
-/*----------------------------------------------------------------*/
-
-#endif
diff --git a/drivers/md/persistent-data/dm-space-map-checker.h b/drivers/md/persistent-data/dm-space-map-checker.h
deleted file mode 100644
index 444dccf6688..00000000000
--- a/drivers/md/persistent-data/dm-space-map-checker.h
+++ /dev/null
@@ -1,26 +0,0 @@
-/*
- * Copyright (C) 2011 Red Hat, Inc.
- *
- * This file is released under the GPL.
- */
-
-#ifndef SNAPSHOTS_SPACE_MAP_CHECKER_H
-#define SNAPSHOTS_SPACE_MAP_CHECKER_H
-
-#include "dm-space-map.h"
-
-/*----------------------------------------------------------------*/
-
-/*
- * This space map wraps a real on-disk space map, and verifies all of its
- * operations.  It uses a lot of memory, so only use if you have a specific
- * problem that you're debugging.
- *
- * Ownership of @sm passes.
- */
-struct dm_space_map *dm_sm_checker_create(struct dm_space_map *sm);
-struct dm_space_map *dm_sm_checker_create_fresh(struct dm_space_map *sm);
-
-/*----------------------------------------------------------------*/
-
-#endif
diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c
index ff3beed6ad2..d77602d63c8 100644
--- a/drivers/md/persistent-data/dm-space-map-common.c
+++ b/drivers/md/persistent-data/dm-space-map-common.c
@@ -224,6 +224,7 @@ static int sm_ll_init(struct ll_disk *ll, struct dm_transaction_manager *tm)
 	ll->nr_blocks = 0;
 	ll->bitmap_root = 0;
 	ll->ref_count_root = 0;
+	ll->bitmap_index_changed = false;
 
 	return 0;
 }
@@ -476,7 +477,15 @@ int sm_ll_dec(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev)
 
 int sm_ll_commit(struct ll_disk *ll)
 {
-	return ll->commit(ll);
+	int r = 0;
+
+	if (ll->bitmap_index_changed) {
+		r = ll->commit(ll);
+		if (!r)
+			ll->bitmap_index_changed = false;
+	}
+
+	return r;
 }
 
 /*----------------------------------------------------------------*/
@@ -491,6 +500,7 @@ static int metadata_ll_load_ie(struct ll_disk *ll, dm_block_t index,
 static int metadata_ll_save_ie(struct ll_disk *ll, dm_block_t index,
 			       struct disk_index_entry *ie)
 {
+	ll->bitmap_index_changed = true;
 	memcpy(ll->mi_le.index + index, ie, sizeof(*ie));
 	return 0;
 }
diff --git a/drivers/md/persistent-data/dm-space-map-common.h b/drivers/md/persistent-data/dm-space-map-common.h
index 8f220821a9a..b3078d5eda0 100644
--- a/drivers/md/persistent-data/dm-space-map-common.h
+++ b/drivers/md/persistent-data/dm-space-map-common.h
@@ -78,6 +78,7 @@ struct ll_disk {
 	open_index_fn open_index;
 	max_index_entries_fn max_entries;
 	commit_fn commit;
+	bool bitmap_index_changed:1;
 };
 
 struct disk_sm_root {
diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c
index fc469ba9f62..f6d29e614ab 100644
--- a/drivers/md/persistent-data/dm-space-map-disk.c
+++ b/drivers/md/persistent-data/dm-space-map-disk.c
@@ -4,7 +4,6 @@
  * This file is released under the GPL.
  */
 
-#include "dm-space-map-checker.h"
 #include "dm-space-map-common.h"
 #include "dm-space-map-disk.h"
 #include "dm-space-map.h"
@@ -252,9 +251,8 @@ static struct dm_space_map ops = {
 	.copy_root = sm_disk_copy_root
 };
 
-static struct dm_space_map *dm_sm_disk_create_real(
-	struct dm_transaction_manager *tm,
-	dm_block_t nr_blocks)
+struct dm_space_map *dm_sm_disk_create(struct dm_transaction_manager *tm,
+				       dm_block_t nr_blocks)
 {
 	int r;
 	struct sm_disk *smd;
@@ -285,18 +283,10 @@ bad:
 	kfree(smd);
 	return ERR_PTR(r);
 }
-
-struct dm_space_map *dm_sm_disk_create(struct dm_transaction_manager *tm,
-				       dm_block_t nr_blocks)
-{
-	struct dm_space_map *sm = dm_sm_disk_create_real(tm, nr_blocks);
-	return dm_sm_checker_create_fresh(sm);
-}
 EXPORT_SYMBOL_GPL(dm_sm_disk_create);
 
-static struct dm_space_map *dm_sm_disk_open_real(
-	struct dm_transaction_manager *tm,
-	void *root_le, size_t len)
+struct dm_space_map *dm_sm_disk_open(struct dm_transaction_manager *tm,
+				     void *root_le, size_t len)
 {
 	int r;
 	struct sm_disk *smd;
@@ -323,13 +313,6 @@ bad:
 	kfree(smd);
 	return ERR_PTR(r);
 }
-
-struct dm_space_map *dm_sm_disk_open(struct dm_transaction_manager *tm,
-				     void *root_le, size_t len)
-{
-	return dm_sm_checker_create(
-		dm_sm_disk_open_real(tm, root_le, len));
-}
 EXPORT_SYMBOL_GPL(dm_sm_disk_open);
 
 /*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-transaction-manager.c b/drivers/md/persistent-data/dm-transaction-manager.c
index 400fe144c0c..d247a35da3c 100644
--- a/drivers/md/persistent-data/dm-transaction-manager.c
+++ b/drivers/md/persistent-data/dm-transaction-manager.c
@@ -5,7 +5,6 @@
  */
 #include "dm-transaction-manager.h"
 #include "dm-space-map.h"
-#include "dm-space-map-checker.h"
 #include "dm-space-map-disk.h"
 #include "dm-space-map-metadata.h"
 #include "dm-persistent-data-internal.h"
@@ -138,6 +137,9 @@ EXPORT_SYMBOL_GPL(dm_tm_create_non_blocking_clone);
 
 void dm_tm_destroy(struct dm_transaction_manager *tm)
 {
+	if (!tm->is_clone)
+		wipe_shadow_table(tm);
+
 	kfree(tm);
 }
 EXPORT_SYMBOL_GPL(dm_tm_destroy);
@@ -217,13 +219,24 @@ static int __shadow_block(struct dm_transaction_manager *tm, dm_block_t orig,
 	if (r < 0)
 		return r;
 
-	r = dm_bm_unlock_move(orig_block, new);
-	if (r < 0) {
+	/*
+	 * It would be tempting to use dm_bm_unlock_move here, but some
+	 * code, such as the space maps, keeps using the old data structures
+	 * secure in the knowledge they won't be changed until the next
+	 * transaction.  Using unlock_move would force a synchronous read
+	 * since the old block would no longer be in the cache.
+	 */
+	r = dm_bm_write_lock_zero(tm->bm, new, v, result);
+	if (r) {
 		dm_bm_unlock(orig_block);
 		return r;
 	}
 
-	return dm_bm_write_lock(tm->bm, new, v, result);
+	memcpy(dm_block_data(*result), dm_block_data(orig_block),
+	       dm_bm_block_size(tm->bm));
+
+	dm_bm_unlock(orig_block);
+	return r;
 }
 
 int dm_tm_shadow_block(struct dm_transaction_manager *tm, dm_block_t orig,
@@ -308,94 +321,61 @@ struct dm_block_manager *dm_tm_get_bm(struct dm_transaction_manager *tm)
 
 static int dm_tm_create_internal(struct dm_block_manager *bm,
 				 dm_block_t sb_location,
-				 struct dm_block_validator *sb_validator,
-				 size_t root_offset, size_t root_max_len,
 				 struct dm_transaction_manager **tm,
 				 struct dm_space_map **sm,
-				 struct dm_block **sblock,
-				 int create)
+				 int create,
+				 void *sm_root, size_t sm_len)
 {
 	int r;
-	struct dm_space_map *inner;
 
-	inner = dm_sm_metadata_init();
-	if (IS_ERR(inner))
-		return PTR_ERR(inner);
+	*sm = dm_sm_metadata_init();
+	if (IS_ERR(*sm))
+		return PTR_ERR(*sm);
 
-	*tm = dm_tm_create(bm, inner);
+	*tm = dm_tm_create(bm, *sm);
 	if (IS_ERR(*tm)) {
-		dm_sm_destroy(inner);
+		dm_sm_destroy(*sm);
 		return PTR_ERR(*tm);
 	}
 
 	if (create) {
-		r = dm_bm_write_lock_zero(dm_tm_get_bm(*tm), sb_location,
-					  sb_validator, sblock);
-		if (r < 0) {
-			DMERR("couldn't lock superblock");
-			goto bad1;
-		}
-
-		r = dm_sm_metadata_create(inner, *tm, dm_bm_nr_blocks(bm),
+		r = dm_sm_metadata_create(*sm, *tm, dm_bm_nr_blocks(bm),
 					  sb_location);
 		if (r) {
 			DMERR("couldn't create metadata space map");
-			goto bad2;
+			goto bad;
 		}
 
-		*sm = dm_sm_checker_create(inner);
-		if (!*sm)
-			goto bad2;
-
 	} else {
-		r = dm_bm_write_lock(dm_tm_get_bm(*tm), sb_location,
-				     sb_validator, sblock);
-		if (r < 0) {
-			DMERR("couldn't lock superblock");
-			goto bad1;
-		}
-
-		r = dm_sm_metadata_open(inner, *tm,
-					dm_block_data(*sblock) + root_offset,
-					root_max_len);
+		r = dm_sm_metadata_open(*sm, *tm, sm_root, sm_len);
 		if (r) {
 			DMERR("couldn't open metadata space map");
-			goto bad2;
+			goto bad;
 		}
-
-		*sm = dm_sm_checker_create(inner);
-		if (!*sm)
-			goto bad2;
 	}
 
 	return 0;
 
-bad2:
-	dm_tm_unlock(*tm, *sblock);
-bad1:
+bad:
 	dm_tm_destroy(*tm);
-	dm_sm_destroy(inner);
+	dm_sm_destroy(*sm);
 	return r;
 }
 
 int dm_tm_create_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
-			 struct dm_block_validator *sb_validator,
 			 struct dm_transaction_manager **tm,
-			 struct dm_space_map **sm, struct dm_block **sblock)
+			 struct dm_space_map **sm)
 {
-	return dm_tm_create_internal(bm, sb_location, sb_validator,
-				     0, 0, tm, sm, sblock, 1);
+	return dm_tm_create_internal(bm, sb_location, tm, sm, 1, NULL, 0);
 }
 EXPORT_SYMBOL_GPL(dm_tm_create_with_sm);
 
 int dm_tm_open_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
-		       struct dm_block_validator *sb_validator,
-		       size_t root_offset, size_t root_max_len,
+		       void *sm_root, size_t root_len,
 		       struct dm_transaction_manager **tm,
-		       struct dm_space_map **sm, struct dm_block **sblock)
+		       struct dm_space_map **sm)
 {
-	return dm_tm_create_internal(bm, sb_location, sb_validator, root_offset,
-				     root_max_len, tm, sm, sblock, 0);
+	return dm_tm_create_internal(bm, sb_location, tm, sm, 0, sm_root, root_len);
 }
 EXPORT_SYMBOL_GPL(dm_tm_open_with_sm);
 
diff --git a/drivers/md/persistent-data/dm-transaction-manager.h b/drivers/md/persistent-data/dm-transaction-manager.h
index 6da784871db..b5b139076ca 100644
--- a/drivers/md/persistent-data/dm-transaction-manager.h
+++ b/drivers/md/persistent-data/dm-transaction-manager.h
@@ -115,16 +115,17 @@ struct dm_block_manager *dm_tm_get_bm(struct dm_transaction_manager *tm);
  *
  * Returns a tm that has an open transaction to write the new disk sm.
  * Caller should store the new sm root and commit.
+ *
+ * The superblock location is passed so the metadata space map knows it
+ * shouldn't be used.
  */
 int dm_tm_create_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
-			 struct dm_block_validator *sb_validator,
 			 struct dm_transaction_manager **tm,
-			 struct dm_space_map **sm, struct dm_block **sblock);
+			 struct dm_space_map **sm);
 
 int dm_tm_open_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
-		       struct dm_block_validator *sb_validator,
-		       size_t root_offset, size_t root_max_len,
+		       void *sm_root, size_t root_len,
 		       struct dm_transaction_manager **tm,
-		       struct dm_space_map **sm, struct dm_block **sblock);
+		       struct dm_space_map **sm);
 
 #endif	/* _LINUX_DM_TRANSACTION_MANAGER_H */
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index a9c7981ddd2..611b5f79761 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -46,6 +46,20 @@
  */
 #define	NR_RAID1_BIOS 256
 
+/* when we get a read error on a read-only array, we redirect to another
+ * device without failing the first device, or trying to over-write to
+ * correct the read error.  To keep track of bad blocks on a per-bio
+ * level, we store IO_BLOCKED in the appropriate 'bios' pointer
+ */
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting devs[n].bio to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
+
 /* When there are this many requests queue to be written by
  * the raid1 thread, we become 'congested' to provide back-pressure
  * for writeback.
@@ -483,12 +497,14 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 	const sector_t this_sector = r1_bio->sector;
 	int sectors;
 	int best_good_sectors;
-	int start_disk;
-	int best_disk;
-	int i;
+	int best_disk, best_dist_disk, best_pending_disk;
+	int has_nonrot_disk;
+	int disk;
 	sector_t best_dist;
+	unsigned int min_pending;
 	struct md_rdev *rdev;
 	int choose_first;
+	int choose_next_idle;
 
 	rcu_read_lock();
 	/*
@@ -499,26 +515,26 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
  retry:
 	sectors = r1_bio->sectors;
 	best_disk = -1;
+	best_dist_disk = -1;
 	best_dist = MaxSector;
+	best_pending_disk = -1;
+	min_pending = UINT_MAX;
 	best_good_sectors = 0;
+	has_nonrot_disk = 0;
+	choose_next_idle = 0;
 
 	if (conf->mddev->recovery_cp < MaxSector &&
-	    (this_sector + sectors >= conf->next_resync)) {
+	    (this_sector + sectors >= conf->next_resync))
 		choose_first = 1;
-		start_disk = 0;
-	} else {
+	else
 		choose_first = 0;
-		start_disk = conf->last_used;
-	}
 
-	for (i = 0 ; i < conf->raid_disks * 2 ; i++) {
+	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
 		sector_t dist;
 		sector_t first_bad;
 		int bad_sectors;
-
-		int disk = start_disk + i;
-		if (disk >= conf->raid_disks)
-			disk -= conf->raid_disks;
+		unsigned int pending;
+		bool nonrot;
 
 		rdev = rcu_dereference(conf->mirrors[disk].rdev);
 		if (r1_bio->bios[disk] == IO_BLOCKED
@@ -577,22 +593,77 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 		} else
 			best_good_sectors = sectors;
 
+		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
+		has_nonrot_disk |= nonrot;
+		pending = atomic_read(&rdev->nr_pending);
 		dist = abs(this_sector - conf->mirrors[disk].head_position);
-		if (choose_first
-		    /* Don't change to another disk for sequential reads */
-		    || conf->next_seq_sect == this_sector
-		    || dist == 0
-		    /* If device is idle, use it */
-		    || atomic_read(&rdev->nr_pending) == 0) {
+		if (choose_first) {
 			best_disk = disk;
 			break;
 		}
+		/* Don't change to another disk for sequential reads */
+		if (conf->mirrors[disk].next_seq_sect == this_sector
+		    || dist == 0) {
+			int opt_iosize = bdev_io_opt(rdev->bdev) >> 9;
+			struct raid1_info *mirror = &conf->mirrors[disk];
+
+			best_disk = disk;
+			/*
+			 * If buffered sequential IO size exceeds optimal
+			 * iosize, check if there is idle disk. If yes, choose
+			 * the idle disk. read_balance could already choose an
+			 * idle disk before noticing it's a sequential IO in
+			 * this disk. This doesn't matter because this disk
+			 * will idle, next time it will be utilized after the
+			 * first disk has IO size exceeds optimal iosize. In
+			 * this way, iosize of the first disk will be optimal
+			 * iosize at least. iosize of the second disk might be
+			 * small, but not a big deal since when the second disk
+			 * starts IO, the first disk is likely still busy.
+			 */
+			if (nonrot && opt_iosize > 0 &&
+			    mirror->seq_start != MaxSector &&
+			    mirror->next_seq_sect > opt_iosize &&
+			    mirror->next_seq_sect - opt_iosize >=
+			    mirror->seq_start) {
+				choose_next_idle = 1;
+				continue;
+			}
+			break;
+		}
+		/* If device is idle, use it */
+		if (pending == 0) {
+			best_disk = disk;
+			break;
+		}
+
+		if (choose_next_idle)
+			continue;
+
+		if (min_pending > pending) {
+			min_pending = pending;
+			best_pending_disk = disk;
+		}
+
 		if (dist < best_dist) {
 			best_dist = dist;
-			best_disk = disk;
+			best_dist_disk = disk;
 		}
 	}
 
+	/*
+	 * If all disks are rotational, choose the closest disk. If any disk is
+	 * non-rotational, choose the disk with less pending request even the
+	 * disk is rotational, which might/might not be optimal for raids with
+	 * mixed ratation/non-rotational disks depending on workload.
+	 */
+	if (best_disk == -1) {
+		if (has_nonrot_disk)
+			best_disk = best_pending_disk;
+		else
+			best_disk = best_dist_disk;
+	}
+
 	if (best_disk >= 0) {
 		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
 		if (!rdev)
@@ -606,8 +677,11 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 			goto retry;
 		}
 		sectors = best_good_sectors;
-		conf->next_seq_sect = this_sector + sectors;
-		conf->last_used = best_disk;
+
+		if (conf->mirrors[best_disk].next_seq_sect != this_sector)
+			conf->mirrors[best_disk].seq_start = this_sector;
+
+		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
 	}
 	rcu_read_unlock();
 	*max_sectors = sectors;
@@ -870,10 +944,48 @@ do_sync_io:
 	pr_debug("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
 }
 
+struct raid1_plug_cb {
+	struct blk_plug_cb	cb;
+	struct bio_list		pending;
+	int			pending_cnt;
+};
+
+static void raid1_unplug(struct blk_plug_cb *cb, bool from_schedule)
+{
+	struct raid1_plug_cb *plug = container_of(cb, struct raid1_plug_cb,
+						  cb);
+	struct mddev *mddev = plug->cb.data;
+	struct r1conf *conf = mddev->private;
+	struct bio *bio;
+
+	if (from_schedule) {
+		spin_lock_irq(&conf->device_lock);
+		bio_list_merge(&conf->pending_bio_list, &plug->pending);
+		conf->pending_count += plug->pending_cnt;
+		spin_unlock_irq(&conf->device_lock);
+		md_wakeup_thread(mddev->thread);
+		kfree(plug);
+		return;
+	}
+
+	/* we aren't scheduling, so we can do the write-out directly. */
+	bio = bio_list_get(&plug->pending);
+	bitmap_unplug(mddev->bitmap);
+	wake_up(&conf->wait_barrier);
+
+	while (bio) { /* submit pending writes */
+		struct bio *next = bio->bi_next;
+		bio->bi_next = NULL;
+		generic_make_request(bio);
+		bio = next;
+	}
+	kfree(plug);
+}
+
 static void make_request(struct mddev *mddev, struct bio * bio)
 {
 	struct r1conf *conf = mddev->private;
-	struct mirror_info *mirror;
+	struct raid1_info *mirror;
 	struct r1bio *r1_bio;
 	struct bio *read_bio;
 	int i, disks;
@@ -883,7 +995,8 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
 	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
 	struct md_rdev *blocked_rdev;
-	int plugged;
+	struct blk_plug_cb *cb;
+	struct raid1_plug_cb *plug = NULL;
 	int first_clone;
 	int sectors_handled;
 	int max_sectors;
@@ -1034,7 +1147,6 @@ read_again:
 	 * the bad blocks.  Each set of writes gets it's own r1bio
 	 * with a set of bios attached.
 	 */
-	plugged = mddev_check_plugged(mddev);
 
 	disks = conf->raid_disks * 2;
  retry_write:
@@ -1187,10 +1299,23 @@ read_again:
 		mbio->bi_private = r1_bio;
 
 		atomic_inc(&r1_bio->remaining);
+
+		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
+		if (cb)
+			plug = container_of(cb, struct raid1_plug_cb, cb);
+		else
+			plug = NULL;
 		spin_lock_irqsave(&conf->device_lock, flags);
-		bio_list_add(&conf->pending_bio_list, mbio);
-		conf->pending_count++;
+		if (plug) {
+			bio_list_add(&plug->pending, mbio);
+			plug->pending_cnt++;
+		} else {
+			bio_list_add(&conf->pending_bio_list, mbio);
+			conf->pending_count++;
+		}
 		spin_unlock_irqrestore(&conf->device_lock, flags);
+		if (!plug)
+			md_wakeup_thread(mddev->thread);
 	}
 	/* Mustn't call r1_bio_write_done before this next test,
 	 * as it could result in the bio being freed.
@@ -1213,9 +1338,6 @@ read_again:
 
 	/* In case raid1d snuck in to freeze_array */
 	wake_up(&conf->wait_barrier);
-
-	if (do_sync || !bitmap || !plugged)
-		md_wakeup_thread(mddev->thread);
 }
 
 static void status(struct seq_file *seq, struct mddev *mddev)
@@ -1367,7 +1489,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	struct r1conf *conf = mddev->private;
 	int err = -EEXIST;
 	int mirror = 0;
-	struct mirror_info *p;
+	struct raid1_info *p;
 	int first = 0;
 	int last = conf->raid_disks - 1;
 	struct request_queue *q = bdev_get_queue(rdev->bdev);
@@ -1436,7 +1558,7 @@ static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 	struct r1conf *conf = mddev->private;
 	int err = 0;
 	int number = rdev->raid_disk;
-	struct mirror_info *p = conf->mirrors+ number;
+	struct raid1_info *p = conf->mirrors + number;
 
 	if (rdev != p->rdev)
 		p = conf->mirrors + conf->raid_disks + number;
@@ -1821,8 +1943,14 @@ static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
 
 	if (atomic_dec_and_test(&r1_bio->remaining)) {
 		/* if we're here, all write(s) have completed, so clean up */
-		md_done_sync(mddev, r1_bio->sectors, 1);
-		put_buf(r1_bio);
+		int s = r1_bio->sectors;
+		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
+		    test_bit(R1BIO_WriteError, &r1_bio->state))
+			reschedule_retry(r1_bio);
+		else {
+			put_buf(r1_bio);
+			md_done_sync(mddev, s, 1);
+		}
 	}
 }
 
@@ -2170,8 +2298,7 @@ static void raid1d(struct mddev *mddev)
 	blk_start_plug(&plug);
 	for (;;) {
 
-		if (atomic_read(&mddev->plug_cnt) == 0)
-			flush_pending_writes(conf);
+		flush_pending_writes(conf);
 
 		spin_lock_irqsave(&conf->device_lock, flags);
 		if (list_empty(head)) {
@@ -2368,6 +2495,18 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 				bio->bi_rw = READ;
 				bio->bi_end_io = end_sync_read;
 				read_targets++;
+			} else if (!test_bit(WriteErrorSeen, &rdev->flags) &&
+				test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
+				!test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
+				/*
+				 * The device is suitable for reading (InSync),
+				 * but has bad block(s) here. Let's try to correct them,
+				 * if we are doing resync or repair. Otherwise, leave
+				 * this device alone for this sync request.
+				 */
+				bio->bi_rw = WRITE;
+				bio->bi_end_io = end_sync_write;
+				write_targets++;
 			}
 		}
 		if (bio->bi_end_io) {
@@ -2425,7 +2564,10 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 		/* There is nowhere to write, so all non-sync
 		 * drives must be failed - so we are finished
 		 */
-		sector_t rv = max_sector - sector_nr;
+		sector_t rv;
+		if (min_bad > 0)
+			max_sector = sector_nr + min_bad;
+		rv = max_sector - sector_nr;
 		*skipped = 1;
 		put_buf(r1_bio);
 		return rv;
@@ -2488,9 +2630,10 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 	 */
 	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 		atomic_set(&r1_bio->remaining, read_targets);
-		for (i = 0; i < conf->raid_disks * 2; i++) {
+		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
 			bio = r1_bio->bios[i];
 			if (bio->bi_end_io == end_sync_read) {
+				read_targets--;
 				md_sync_acct(bio->bi_bdev, nr_sectors);
 				generic_make_request(bio);
 			}
@@ -2517,7 +2660,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 {
 	struct r1conf *conf;
 	int i;
-	struct mirror_info *disk;
+	struct raid1_info *disk;
 	struct md_rdev *rdev;
 	int err = -ENOMEM;
 
@@ -2525,7 +2668,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	if (!conf)
 		goto abort;
 
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)
+	conf->mirrors = kzalloc(sizeof(struct raid1_info)
 				* mddev->raid_disks * 2,
 				 GFP_KERNEL);
 	if (!conf->mirrors)
@@ -2568,6 +2711,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 			mddev->merge_check_needed = 1;
 
 		disk->head_position = 0;
+		disk->seq_start = MaxSector;
 	}
 	conf->raid_disks = mddev->raid_disks;
 	conf->mddev = mddev;
@@ -2581,7 +2725,6 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	conf->recovery_disabled = mddev->recovery_disabled - 1;
 
 	err = -EIO;
-	conf->last_used = -1;
 	for (i = 0; i < conf->raid_disks * 2; i++) {
 
 		disk = conf->mirrors + i;
@@ -2607,21 +2750,11 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 			if (disk->rdev &&
 			    (disk->rdev->saved_raid_disk < 0))
 				conf->fullsync = 1;
-		} else if (conf->last_used < 0)
-			/*
-			 * The first working device is used as a
-			 * starting point to read balancing.
-			 */
-			conf->last_used = i;
+		}
 	}
 
-	if (conf->last_used < 0) {
-		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
-		       mdname(mddev));
-		goto abort;
-	}
 	err = -ENOMEM;
-	conf->thread = md_register_thread(raid1d, mddev, NULL);
+	conf->thread = md_register_thread(raid1d, mddev, "raid1");
 	if (!conf->thread) {
 		printk(KERN_ERR
 		       "md/raid1:%s: couldn't allocate thread\n",
@@ -2794,7 +2927,7 @@ static int raid1_reshape(struct mddev *mddev)
 	 */
 	mempool_t *newpool, *oldpool;
 	struct pool_info *newpoolinfo;
-	struct mirror_info *newmirrors;
+	struct raid1_info *newmirrors;
 	struct r1conf *conf = mddev->private;
 	int cnt, raid_disks;
 	unsigned long flags;
@@ -2837,7 +2970,7 @@ static int raid1_reshape(struct mddev *mddev)
 		kfree(newpoolinfo);
 		return -ENOMEM;
 	}
-	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks * 2,
+	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
 			     GFP_KERNEL);
 	if (!newmirrors) {
 		kfree(newpoolinfo);
@@ -2876,7 +3009,6 @@ static int raid1_reshape(struct mddev *mddev)
 	conf->raid_disks = mddev->raid_disks = raid_disks;
 	mddev->delta_disks = 0;
 
-	conf->last_used = 0; /* just make sure it is in-range */
 	lower_barrier(conf);
 
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
diff --git a/drivers/md/raid1.h b/drivers/md/raid1.h
index 80ded139314..0ff3715fb7e 100644
--- a/drivers/md/raid1.h
+++ b/drivers/md/raid1.h
@@ -1,9 +1,15 @@
 #ifndef _RAID1_H
 #define _RAID1_H
 
-struct mirror_info {
+struct raid1_info {
 	struct md_rdev	*rdev;
 	sector_t	head_position;
+
+	/* When choose the best device for a read (read_balance())
+	 * we try to keep sequential reads one the same device
+	 */
+	sector_t	next_seq_sect;
+	sector_t	seq_start;
 };
 
 /*
@@ -24,17 +30,11 @@ struct pool_info {
 
 struct r1conf {
 	struct mddev		*mddev;
-	struct mirror_info	*mirrors;	/* twice 'raid_disks' to
+	struct raid1_info	*mirrors;	/* twice 'raid_disks' to
 						 * allow for replacements.
 						 */
 	int			raid_disks;
 
-	/* When choose the best device for a read (read_balance())
-	 * we try to keep sequential reads one the same device
-	 * using 'last_used' and 'next_seq_sect'
-	 */
-	int			last_used;
-	sector_t		next_seq_sect;
 	/* During resync, read_balancing is only allowed on the part
 	 * of the array that has been resynced.  'next_resync' tells us
 	 * where that is.
@@ -135,20 +135,6 @@ struct r1bio {
 	/* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
 };
 
-/* when we get a read error on a read-only array, we redirect to another
- * device without failing the first device, or trying to over-write to
- * correct the read error.  To keep track of bad blocks on a per-bio
- * level, we store IO_BLOCKED in the appropriate 'bios' pointer
- */
-#define IO_BLOCKED ((struct bio *)1)
-/* When we successfully write to a known bad-block, we need to remove the
- * bad-block marking which must be done from process context.  So we record
- * the success by setting bios[n] to IO_MADE_GOOD
- */
-#define IO_MADE_GOOD ((struct bio *)2)
-
-#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
-
 /* bits for r1bio.state */
 #define	R1BIO_Uptodate	0
 #define	R1BIO_IsSync	1
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 99ae6068e45..de5ed6fd880 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -60,7 +60,21 @@
  */
 #define	NR_RAID10_BIOS 256
 
-/* When there are this many requests queue to be written by
+/* when we get a read error on a read-only array, we redirect to another
+ * device without failing the first device, or trying to over-write to
+ * correct the read error.  To keep track of bad blocks on a per-bio
+ * level, we store IO_BLOCKED in the appropriate 'bios' pointer
+ */
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting devs[n].bio to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
+
+/* When there are this many requests queued to be written by
  * the raid10 thread, we become 'congested' to provide back-pressure
  * for writeback.
  */
@@ -717,7 +731,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 	int sectors = r10_bio->sectors;
 	int best_good_sectors;
 	sector_t new_distance, best_dist;
-	struct md_rdev *rdev, *best_rdev;
+	struct md_rdev *best_rdev, *rdev = NULL;
 	int do_balance;
 	int best_slot;
 	struct geom *geo = &conf->geo;
@@ -839,9 +853,8 @@ retry:
 	return rdev;
 }
 
-static int raid10_congested(void *data, int bits)
+int md_raid10_congested(struct mddev *mddev, int bits)
 {
-	struct mddev *mddev = data;
 	struct r10conf *conf = mddev->private;
 	int i, ret = 0;
 
@@ -849,8 +862,6 @@ static int raid10_congested(void *data, int bits)
 	    conf->pending_count >= max_queued_requests)
 		return 1;
 
-	if (mddev_congested(mddev, bits))
-		return 1;
 	rcu_read_lock();
 	for (i = 0;
 	     (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
@@ -866,6 +877,15 @@ static int raid10_congested(void *data, int bits)
 	rcu_read_unlock();
 	return ret;
 }
+EXPORT_SYMBOL_GPL(md_raid10_congested);
+
+static int raid10_congested(void *data, int bits)
+{
+	struct mddev *mddev = data;
+
+	return mddev_congested(mddev, bits) ||
+		md_raid10_congested(mddev, bits);
+}
 
 static void flush_pending_writes(struct r10conf *conf)
 {
@@ -1039,7 +1059,6 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	const unsigned long do_fua = (bio->bi_rw & REQ_FUA);
 	unsigned long flags;
 	struct md_rdev *blocked_rdev;
-	int plugged;
 	int sectors_handled;
 	int max_sectors;
 	int sectors;
@@ -1239,7 +1258,6 @@ read_again:
 	 * of r10_bios is recored in bio->bi_phys_segments just as with
 	 * the read case.
 	 */
-	plugged = mddev_check_plugged(mddev);
 
 	r10_bio->read_slot = -1; /* make sure repl_bio gets freed */
 	raid10_find_phys(conf, r10_bio);
@@ -1396,6 +1414,8 @@ retry_write:
 		bio_list_add(&conf->pending_bio_list, mbio);
 		conf->pending_count++;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
+		if (!mddev_check_plugged(mddev))
+			md_wakeup_thread(mddev->thread);
 
 		if (!r10_bio->devs[i].repl_bio)
 			continue;
@@ -1423,6 +1443,8 @@ retry_write:
 		bio_list_add(&conf->pending_bio_list, mbio);
 		conf->pending_count++;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
+		if (!mddev_check_plugged(mddev))
+			md_wakeup_thread(mddev->thread);
 	}
 
 	/* Don't remove the bias on 'remaining' (one_write_done) until
@@ -1448,9 +1470,6 @@ retry_write:
 
 	/* In case raid10d snuck in to freeze_array */
 	wake_up(&conf->wait_barrier);
-
-	if (do_sync || !mddev->bitmap || !plugged)
-		md_wakeup_thread(mddev->thread);
 }
 
 static void status(struct seq_file *seq, struct mddev *mddev)
@@ -1547,7 +1566,7 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 static void print_conf(struct r10conf *conf)
 {
 	int i;
-	struct mirror_info *tmp;
+	struct raid10_info *tmp;
 
 	printk(KERN_DEBUG "RAID10 conf printout:\n");
 	if (!conf) {
@@ -1581,7 +1600,7 @@ static int raid10_spare_active(struct mddev *mddev)
 {
 	int i;
 	struct r10conf *conf = mddev->private;
-	struct mirror_info *tmp;
+	struct raid10_info *tmp;
 	int count = 0;
 	unsigned long flags;
 
@@ -1656,7 +1675,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	else
 		mirror = first;
 	for ( ; mirror <= last ; mirror++) {
-		struct mirror_info *p = &conf->mirrors[mirror];
+		struct raid10_info *p = &conf->mirrors[mirror];
 		if (p->recovery_disabled == mddev->recovery_disabled)
 			continue;
 		if (p->rdev) {
@@ -1710,7 +1729,7 @@ static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 	int err = 0;
 	int number = rdev->raid_disk;
 	struct md_rdev **rdevp;
-	struct mirror_info *p = conf->mirrors + number;
+	struct raid10_info *p = conf->mirrors + number;
 
 	print_conf(conf);
 	if (rdev == p->rdev)
@@ -2310,7 +2329,7 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 			if (r10_sync_page_io(rdev,
 					     r10_bio->devs[sl].addr +
 					     sect,
-					     s<<9, conf->tmppage, WRITE)
+					     s, conf->tmppage, WRITE)
 			    == 0) {
 				/* Well, this device is dead */
 				printk(KERN_NOTICE
@@ -2349,7 +2368,7 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 			switch (r10_sync_page_io(rdev,
 					     r10_bio->devs[sl].addr +
 					     sect,
-					     s<<9, conf->tmppage,
+					     s, conf->tmppage,
 						 READ)) {
 			case 0:
 				/* Well, this device is dead */
@@ -2512,7 +2531,7 @@ read_more:
 	slot = r10_bio->read_slot;
 	printk_ratelimited(
 		KERN_ERR
-		"md/raid10:%s: %s: redirecting"
+		"md/raid10:%s: %s: redirecting "
 		"sector %llu to another mirror\n",
 		mdname(mddev),
 		bdevname(rdev->bdev, b),
@@ -2876,7 +2895,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			sector_t sect;
 			int must_sync;
 			int any_working;
-			struct mirror_info *mirror = &conf->mirrors[i];
+			struct raid10_info *mirror = &conf->mirrors[i];
 
 			if ((mirror->rdev == NULL ||
 			     test_bit(In_sync, &mirror->rdev->flags))
@@ -2890,6 +2909,12 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			/* want to reconstruct this device */
 			rb2 = r10_bio;
 			sect = raid10_find_virt(conf, sector_nr, i);
+			if (sect >= mddev->resync_max_sectors) {
+				/* last stripe is not complete - don't
+				 * try to recover this sector.
+				 */
+				continue;
+			}
 			/* Unless we are doing a full sync, or a replacement
 			 * we only need to recover the block if it is set in
 			 * the bitmap
@@ -3382,7 +3407,7 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 		goto out;
 
 	/* FIXME calc properly */
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*(mddev->raid_disks +
+	conf->mirrors = kzalloc(sizeof(struct raid10_info)*(mddev->raid_disks +
 							    max(0,mddev->delta_disks)),
 				GFP_KERNEL);
 	if (!conf->mirrors)
@@ -3421,7 +3446,7 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	spin_lock_init(&conf->resync_lock);
 	init_waitqueue_head(&conf->wait_barrier);
 
-	conf->thread = md_register_thread(raid10d, mddev, NULL);
+	conf->thread = md_register_thread(raid10d, mddev, "raid10");
 	if (!conf->thread)
 		goto out;
 
@@ -3446,7 +3471,7 @@ static int run(struct mddev *mddev)
 {
 	struct r10conf *conf;
 	int i, disk_idx, chunk_size;
-	struct mirror_info *disk;
+	struct raid10_info *disk;
 	struct md_rdev *rdev;
 	sector_t size;
 	sector_t min_offset_diff = 0;
@@ -3466,12 +3491,14 @@ static int run(struct mddev *mddev)
 	conf->thread = NULL;
 
 	chunk_size = mddev->chunk_sectors << 9;
-	blk_queue_io_min(mddev->queue, chunk_size);
-	if (conf->geo.raid_disks % conf->geo.near_copies)
-		blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
-	else
-		blk_queue_io_opt(mddev->queue, chunk_size *
-				 (conf->geo.raid_disks / conf->geo.near_copies));
+	if (mddev->queue) {
+		blk_queue_io_min(mddev->queue, chunk_size);
+		if (conf->geo.raid_disks % conf->geo.near_copies)
+			blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
+		else
+			blk_queue_io_opt(mddev->queue, chunk_size *
+					 (conf->geo.raid_disks / conf->geo.near_copies));
+	}
 
 	rdev_for_each(rdev, mddev) {
 		long long diff;
@@ -3505,8 +3532,9 @@ static int run(struct mddev *mddev)
 		if (first || diff < min_offset_diff)
 			min_offset_diff = diff;
 
-		disk_stack_limits(mddev->gendisk, rdev->bdev,
-				  rdev->data_offset << 9);
+		if (mddev->gendisk)
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 
 		disk->head_position = 0;
 	}
@@ -3569,22 +3597,22 @@ static int run(struct mddev *mddev)
 	md_set_array_sectors(mddev, size);
 	mddev->resync_max_sectors = size;
 
-	mddev->queue->backing_dev_info.congested_fn = raid10_congested;
-	mddev->queue->backing_dev_info.congested_data = mddev;
-
-	/* Calculate max read-ahead size.
-	 * We need to readahead at least twice a whole stripe....
-	 * maybe...
-	 */
-	{
+	if (mddev->queue) {
 		int stripe = conf->geo.raid_disks *
 			((mddev->chunk_sectors << 9) / PAGE_SIZE);
+		mddev->queue->backing_dev_info.congested_fn = raid10_congested;
+		mddev->queue->backing_dev_info.congested_data = mddev;
+
+		/* Calculate max read-ahead size.
+		 * We need to readahead at least twice a whole stripe....
+		 * maybe...
+		 */
 		stripe /= conf->geo.near_copies;
 		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
 			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
+		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 	}
 
-	blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 
 	if (md_integrity_register(mddev))
 		goto out_free_conf;
@@ -3635,7 +3663,10 @@ static int stop(struct mddev *mddev)
 	lower_barrier(conf);
 
 	md_unregister_thread(&mddev->thread);
-	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
+	if (mddev->queue)
+		/* the unplug fn references 'conf'*/
+		blk_sync_queue(mddev->queue);
+
 	if (conf->r10bio_pool)
 		mempool_destroy(conf->r10bio_pool);
 	kfree(conf->mirrors);
@@ -3799,7 +3830,7 @@ static int raid10_check_reshape(struct mddev *mddev)
 	if (mddev->delta_disks > 0) {
 		/* allocate new 'mirrors' list */
 		conf->mirrors_new = kzalloc(
-			sizeof(struct mirror_info)
+			sizeof(struct raid10_info)
 			*(mddev->raid_disks +
 			  mddev->delta_disks),
 			GFP_KERNEL);
@@ -3924,7 +3955,7 @@ static int raid10_start_reshape(struct mddev *mddev)
 	spin_lock_irq(&conf->device_lock);
 	if (conf->mirrors_new) {
 		memcpy(conf->mirrors_new, conf->mirrors,
-		       sizeof(struct mirror_info)*conf->prev.raid_disks);
+		       sizeof(struct raid10_info)*conf->prev.raid_disks);
 		smp_mb();
 		kfree(conf->mirrors_old); /* FIXME and elsewhere */
 		conf->mirrors_old = conf->mirrors;
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 135b1b0a155..007c2c68dd8 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -1,7 +1,7 @@
 #ifndef _RAID10_H
 #define _RAID10_H
 
-struct mirror_info {
+struct raid10_info {
 	struct md_rdev	*rdev, *replacement;
 	sector_t	head_position;
 	int		recovery_disabled;	/* matches
@@ -13,8 +13,8 @@ struct mirror_info {
 
 struct r10conf {
 	struct mddev		*mddev;
-	struct mirror_info	*mirrors;
-	struct mirror_info	*mirrors_new, *mirrors_old;
+	struct raid10_info	*mirrors;
+	struct raid10_info	*mirrors_new, *mirrors_old;
 	spinlock_t		device_lock;
 
 	/* geometry */
@@ -123,20 +123,6 @@ struct r10bio {
 	} devs[0];
 };
 
-/* when we get a read error on a read-only array, we redirect to another
- * device without failing the first device, or trying to over-write to
- * correct the read error.  To keep track of bad blocks on a per-bio
- * level, we store IO_BLOCKED in the appropriate 'bios' pointer
- */
-#define IO_BLOCKED ((struct bio*)1)
-/* When we successfully write to a known bad-block, we need to remove the
- * bad-block marking which must be done from process context.  So we record
- * the success by setting devs[n].bio to IO_MADE_GOOD
- */
-#define IO_MADE_GOOD ((struct bio *)2)
-
-#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
-
 /* bits for r10bio.state */
 enum r10bio_state {
 	R10BIO_Uptodate,
@@ -159,4 +145,7 @@ enum r10bio_state {
  */
 	R10BIO_Previous,
 };
+
+extern int md_raid10_congested(struct mddev *mddev, int bits);
+
 #endif
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index d26767246d2..adda94df5eb 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -99,34 +99,40 @@ static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
  * We maintain a biased count of active stripes in the bottom 16 bits of
  * bi_phys_segments, and a count of processed stripes in the upper 16 bits
  */
-static inline int raid5_bi_phys_segments(struct bio *bio)
+static inline int raid5_bi_processed_stripes(struct bio *bio)
 {
-	return bio->bi_phys_segments & 0xffff;
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return (atomic_read(segments) >> 16) & 0xffff;
 }
 
-static inline int raid5_bi_hw_segments(struct bio *bio)
+static inline int raid5_dec_bi_active_stripes(struct bio *bio)
 {
-	return (bio->bi_phys_segments >> 16) & 0xffff;
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return atomic_sub_return(1, segments) & 0xffff;
 }
 
-static inline int raid5_dec_bi_phys_segments(struct bio *bio)
+static inline void raid5_inc_bi_active_stripes(struct bio *bio)
 {
-	--bio->bi_phys_segments;
-	return raid5_bi_phys_segments(bio);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_inc(segments);
 }
 
-static inline int raid5_dec_bi_hw_segments(struct bio *bio)
+static inline void raid5_set_bi_processed_stripes(struct bio *bio,
+	unsigned int cnt)
 {
-	unsigned short val = raid5_bi_hw_segments(bio);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	int old, new;
 
-	--val;
-	bio->bi_phys_segments = (val << 16) | raid5_bi_phys_segments(bio);
-	return val;
+	do {
+		old = atomic_read(segments);
+		new = (old & 0xffff) | (cnt << 16);
+	} while (atomic_cmpxchg(segments, old, new) != old);
 }
 
-static inline void raid5_set_bi_hw_segments(struct bio *bio, unsigned int cnt)
+static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
 {
-	bio->bi_phys_segments = raid5_bi_phys_segments(bio) | (cnt << 16);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_set(segments, cnt);
 }
 
 /* Find first data disk in a raid6 stripe */
@@ -190,47 +196,56 @@ static int stripe_operations_active(struct stripe_head *sh)
 	       test_bit(STRIPE_COMPUTE_RUN, &sh->state);
 }
 
-static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
+static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
 {
-	if (atomic_dec_and_test(&sh->count)) {
-		BUG_ON(!list_empty(&sh->lru));
-		BUG_ON(atomic_read(&conf->active_stripes)==0);
-		if (test_bit(STRIPE_HANDLE, &sh->state)) {
-			if (test_bit(STRIPE_DELAYED, &sh->state))
-				list_add_tail(&sh->lru, &conf->delayed_list);
-			else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
-				   sh->bm_seq - conf->seq_write > 0)
-				list_add_tail(&sh->lru, &conf->bitmap_list);
-			else {
-				clear_bit(STRIPE_BIT_DELAY, &sh->state);
-				list_add_tail(&sh->lru, &conf->handle_list);
-			}
-			md_wakeup_thread(conf->mddev->thread);
-		} else {
-			BUG_ON(stripe_operations_active(sh));
-			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
-				if (atomic_dec_return(&conf->preread_active_stripes)
-				    < IO_THRESHOLD)
-					md_wakeup_thread(conf->mddev->thread);
-			atomic_dec(&conf->active_stripes);
-			if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
-				list_add_tail(&sh->lru, &conf->inactive_list);
-				wake_up(&conf->wait_for_stripe);
-				if (conf->retry_read_aligned)
-					md_wakeup_thread(conf->mddev->thread);
-			}
+	BUG_ON(!list_empty(&sh->lru));
+	BUG_ON(atomic_read(&conf->active_stripes)==0);
+	if (test_bit(STRIPE_HANDLE, &sh->state)) {
+		if (test_bit(STRIPE_DELAYED, &sh->state) &&
+		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			list_add_tail(&sh->lru, &conf->delayed_list);
+		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
+			   sh->bm_seq - conf->seq_write > 0)
+			list_add_tail(&sh->lru, &conf->bitmap_list);
+		else {
+			clear_bit(STRIPE_DELAYED, &sh->state);
+			clear_bit(STRIPE_BIT_DELAY, &sh->state);
+			list_add_tail(&sh->lru, &conf->handle_list);
+		}
+		md_wakeup_thread(conf->mddev->thread);
+	} else {
+		BUG_ON(stripe_operations_active(sh));
+		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			if (atomic_dec_return(&conf->preread_active_stripes)
+			    < IO_THRESHOLD)
+				md_wakeup_thread(conf->mddev->thread);
+		atomic_dec(&conf->active_stripes);
+		if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
+			list_add_tail(&sh->lru, &conf->inactive_list);
+			wake_up(&conf->wait_for_stripe);
+			if (conf->retry_read_aligned)
+				md_wakeup_thread(conf->mddev->thread);
 		}
 	}
 }
 
+static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
+{
+	if (atomic_dec_and_test(&sh->count))
+		do_release_stripe(conf, sh);
+}
+
 static void release_stripe(struct stripe_head *sh)
 {
 	struct r5conf *conf = sh->raid_conf;
 	unsigned long flags;
 
-	spin_lock_irqsave(&conf->device_lock, flags);
-	__release_stripe(conf, sh);
-	spin_unlock_irqrestore(&conf->device_lock, flags);
+	local_irq_save(flags);
+	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
+		do_release_stripe(conf, sh);
+		spin_unlock(&conf->device_lock);
+	}
+	local_irq_restore(flags);
 }
 
 static inline void remove_hash(struct stripe_head *sh)
@@ -469,7 +484,8 @@ get_active_stripe(struct r5conf *conf, sector_t sector,
 		} else {
 			if (atomic_read(&sh->count)) {
 				BUG_ON(!list_empty(&sh->lru)
-				    && !test_bit(STRIPE_EXPANDING, &sh->state));
+				    && !test_bit(STRIPE_EXPANDING, &sh->state)
+				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
 			} else {
 				if (!test_bit(STRIPE_HANDLE, &sh->state))
 					atomic_inc(&conf->active_stripes);
@@ -606,6 +622,12 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 					 * a chance*/
 					md_check_recovery(conf->mddev);
 				}
+				/*
+				 * Because md_wait_for_blocked_rdev
+				 * will dec nr_pending, we must
+				 * increment it first.
+				 */
+				atomic_inc(&rdev->nr_pending);
 				md_wait_for_blocked_rdev(rdev, conf->mddev);
 			} else {
 				/* Acknowledged bad block - skip the write */
@@ -632,6 +654,9 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 			else
 				bi->bi_sector = (sh->sector
 						 + rdev->data_offset);
+			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+				bi->bi_rw |= REQ_FLUSH;
+
 			bi->bi_flags = 1 << BIO_UPTODATE;
 			bi->bi_idx = 0;
 			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
@@ -741,14 +766,12 @@ static void ops_complete_biofill(void *stripe_head_ref)
 {
 	struct stripe_head *sh = stripe_head_ref;
 	struct bio *return_bi = NULL;
-	struct r5conf *conf = sh->raid_conf;
 	int i;
 
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
 
 	/* clear completed biofills */
-	spin_lock_irq(&conf->device_lock);
 	for (i = sh->disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
 
@@ -766,7 +789,7 @@ static void ops_complete_biofill(void *stripe_head_ref)
 			while (rbi && rbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				rbi2 = r5_next_bio(rbi, dev->sector);
-				if (!raid5_dec_bi_phys_segments(rbi)) {
+				if (!raid5_dec_bi_active_stripes(rbi)) {
 					rbi->bi_next = return_bi;
 					return_bi = rbi;
 				}
@@ -774,7 +797,6 @@ static void ops_complete_biofill(void *stripe_head_ref)
 			}
 		}
 	}
-	spin_unlock_irq(&conf->device_lock);
 	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
 
 	return_io(return_bi);
@@ -786,7 +808,6 @@ static void ops_complete_biofill(void *stripe_head_ref)
 static void ops_run_biofill(struct stripe_head *sh)
 {
 	struct dma_async_tx_descriptor *tx = NULL;
-	struct r5conf *conf = sh->raid_conf;
 	struct async_submit_ctl submit;
 	int i;
 
@@ -797,10 +818,10 @@ static void ops_run_biofill(struct stripe_head *sh)
 		struct r5dev *dev = &sh->dev[i];
 		if (test_bit(R5_Wantfill, &dev->flags)) {
 			struct bio *rbi;
-			spin_lock_irq(&conf->device_lock);
+			spin_lock_irq(&sh->stripe_lock);
 			dev->read = rbi = dev->toread;
 			dev->toread = NULL;
-			spin_unlock_irq(&conf->device_lock);
+			spin_unlock_irq(&sh->stripe_lock);
 			while (rbi && rbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				tx = async_copy_data(0, rbi, dev->page,
@@ -1136,12 +1157,12 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
 			struct bio *wbi;
 
-			spin_lock_irq(&sh->raid_conf->device_lock);
+			spin_lock_irq(&sh->stripe_lock);
 			chosen = dev->towrite;
 			dev->towrite = NULL;
 			BUG_ON(dev->written);
 			wbi = dev->written = chosen;
-			spin_unlock_irq(&sh->raid_conf->device_lock);
+			spin_unlock_irq(&sh->stripe_lock);
 
 			while (wbi && wbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
@@ -1446,6 +1467,8 @@ static int grow_one_stripe(struct r5conf *conf)
 	init_waitqueue_head(&sh->ops.wait_for_ops);
 	#endif
 
+	spin_lock_init(&sh->stripe_lock);
+
 	if (grow_buffers(sh)) {
 		shrink_buffers(sh);
 		kmem_cache_free(conf->slab_cache, sh);
@@ -1731,12 +1754,15 @@ static void raid5_end_read_request(struct bio * bi, int error)
 			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
-		}
+		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
+
 		if (atomic_read(&rdev->read_errors))
 			atomic_set(&rdev->read_errors, 0);
 	} else {
 		const char *bdn = bdevname(rdev->bdev, b);
 		int retry = 0;
+		int set_bad = 0;
 
 		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
 		atomic_inc(&rdev->read_errors);
@@ -1748,7 +1774,8 @@ static void raid5_end_read_request(struct bio * bi, int error)
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
-		else if (conf->mddev->degraded >= conf->max_degraded)
+		else if (conf->mddev->degraded >= conf->max_degraded) {
+			set_bad = 1;
 			printk_ratelimited(
 				KERN_WARNING
 				"md/raid:%s: read error not correctable "
@@ -1756,8 +1783,9 @@ static void raid5_end_read_request(struct bio * bi, int error)
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
-		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
+		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
 			/* Oh, no!!! */
+			set_bad = 1;
 			printk_ratelimited(
 				KERN_WARNING
 				"md/raid:%s: read error NOT corrected!! "
@@ -1765,7 +1793,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
-		else if (atomic_read(&rdev->read_errors)
+		} else if (atomic_read(&rdev->read_errors)
 			 > conf->max_nr_stripes)
 			printk(KERN_WARNING
 			       "md/raid:%s: Too many read errors, failing device %s.\n",
@@ -1773,11 +1801,19 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		else
 			retry = 1;
 		if (retry)
-			set_bit(R5_ReadError, &sh->dev[i].flags);
+			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
+				set_bit(R5_ReadError, &sh->dev[i].flags);
+				clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
+			} else
+				set_bit(R5_ReadNoMerge, &sh->dev[i].flags);
 		else {
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
-			md_error(conf->mddev, rdev);
+			if (!(set_bad
+			      && test_bit(In_sync, &rdev->flags)
+			      && rdev_set_badblocks(
+				      rdev, sh->sector, STRIPE_SECTORS, 0)))
+				md_error(conf->mddev, rdev);
 		}
 	}
 	rdev_dec_pending(rdev, conf->mddev);
@@ -2325,11 +2361,18 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 		(unsigned long long)bi->bi_sector,
 		(unsigned long long)sh->sector);
 
-
-	spin_lock_irq(&conf->device_lock);
+	/*
+	 * If several bio share a stripe. The bio bi_phys_segments acts as a
+	 * reference count to avoid race. The reference count should already be
+	 * increased before this function is called (for example, in
+	 * make_request()), so other bio sharing this stripe will not free the
+	 * stripe. If a stripe is owned by one stripe, the stripe lock will
+	 * protect it.
+	 */
+	spin_lock_irq(&sh->stripe_lock);
 	if (forwrite) {
 		bip = &sh->dev[dd_idx].towrite;
-		if (*bip == NULL && sh->dev[dd_idx].written == NULL)
+		if (*bip == NULL)
 			firstwrite = 1;
 	} else
 		bip = &sh->dev[dd_idx].toread;
@@ -2345,7 +2388,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 	if (*bip)
 		bi->bi_next = *bip;
 	*bip = bi;
-	bi->bi_phys_segments++;
+	raid5_inc_bi_active_stripes(bi);
 
 	if (forwrite) {
 		/* check if page is covered */
@@ -2360,7 +2403,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
 			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
 	}
-	spin_unlock_irq(&conf->device_lock);
+	spin_unlock_irq(&sh->stripe_lock);
 
 	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
 		(unsigned long long)(*bip)->bi_sector,
@@ -2376,7 +2419,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 
  overlap:
 	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
-	spin_unlock_irq(&conf->device_lock);
+	spin_unlock_irq(&sh->stripe_lock);
 	return 0;
 }
 
@@ -2426,10 +2469,11 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 				rdev_dec_pending(rdev, conf->mddev);
 			}
 		}
-		spin_lock_irq(&conf->device_lock);
+		spin_lock_irq(&sh->stripe_lock);
 		/* fail all writes first */
 		bi = sh->dev[i].towrite;
 		sh->dev[i].towrite = NULL;
+		spin_unlock_irq(&sh->stripe_lock);
 		if (bi) {
 			s->to_write--;
 			bitmap_end = 1;
@@ -2442,13 +2486,17 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 			sh->dev[i].sector + STRIPE_SECTORS) {
 			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
-			if (!raid5_dec_bi_phys_segments(bi)) {
+			if (!raid5_dec_bi_active_stripes(bi)) {
 				md_write_end(conf->mddev);
 				bi->bi_next = *return_bi;
 				*return_bi = bi;
 			}
 			bi = nextbi;
 		}
+		if (bitmap_end)
+			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+				STRIPE_SECTORS, 0, 0);
+		bitmap_end = 0;
 		/* and fail all 'written' */
 		bi = sh->dev[i].written;
 		sh->dev[i].written = NULL;
@@ -2457,7 +2505,7 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 		       sh->dev[i].sector + STRIPE_SECTORS) {
 			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
-			if (!raid5_dec_bi_phys_segments(bi)) {
+			if (!raid5_dec_bi_active_stripes(bi)) {
 				md_write_end(conf->mddev);
 				bi->bi_next = *return_bi;
 				*return_bi = bi;
@@ -2481,14 +2529,13 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 				struct bio *nextbi =
 					r5_next_bio(bi, sh->dev[i].sector);
 				clear_bit(BIO_UPTODATE, &bi->bi_flags);
-				if (!raid5_dec_bi_phys_segments(bi)) {
+				if (!raid5_dec_bi_active_stripes(bi)) {
 					bi->bi_next = *return_bi;
 					*return_bi = bi;
 				}
 				bi = nextbi;
 			}
 		}
-		spin_unlock_irq(&conf->device_lock);
 		if (bitmap_end)
 			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
 					STRIPE_SECTORS, 0, 0);
@@ -2692,30 +2739,23 @@ static void handle_stripe_clean_event(struct r5conf *conf,
 				test_bit(R5_UPTODATE, &dev->flags)) {
 				/* We can return any write requests */
 				struct bio *wbi, *wbi2;
-				int bitmap_end = 0;
 				pr_debug("Return write for disc %d\n", i);
-				spin_lock_irq(&conf->device_lock);
 				wbi = dev->written;
 				dev->written = NULL;
 				while (wbi && wbi->bi_sector <
 					dev->sector + STRIPE_SECTORS) {
 					wbi2 = r5_next_bio(wbi, dev->sector);
-					if (!raid5_dec_bi_phys_segments(wbi)) {
+					if (!raid5_dec_bi_active_stripes(wbi)) {
 						md_write_end(conf->mddev);
 						wbi->bi_next = *return_bi;
 						*return_bi = wbi;
 					}
 					wbi = wbi2;
 				}
-				if (dev->towrite == NULL)
-					bitmap_end = 1;
-				spin_unlock_irq(&conf->device_lock);
-				if (bitmap_end)
-					bitmap_endwrite(conf->mddev->bitmap,
-							sh->sector,
-							STRIPE_SECTORS,
+				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+						STRIPE_SECTORS,
 					 !test_bit(STRIPE_DEGRADED, &sh->state),
-							0);
+						0);
 			}
 		}
 
@@ -3167,7 +3207,6 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 
 	/* Now to look around and see what can be done */
 	rcu_read_lock();
-	spin_lock_irq(&conf->device_lock);
 	for (i=disks; i--; ) {
 		struct md_rdev *rdev;
 		sector_t first_bad;
@@ -3313,7 +3352,6 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 				do_recovery = 1;
 		}
 	}
-	spin_unlock_irq(&conf->device_lock);
 	if (test_bit(STRIPE_SYNCING, &sh->state)) {
 		/* If there is a failed device being replaced,
 		 *     we must be recovering.
@@ -3582,8 +3620,18 @@ static void handle_stripe(struct stripe_head *sh)
 
 finish:
 	/* wait for this device to become unblocked */
-	if (conf->mddev->external && unlikely(s.blocked_rdev))
-		md_wait_for_blocked_rdev(s.blocked_rdev, conf->mddev);
+	if (unlikely(s.blocked_rdev)) {
+		if (conf->mddev->external)
+			md_wait_for_blocked_rdev(s.blocked_rdev,
+						 conf->mddev);
+		else
+			/* Internal metadata will immediately
+			 * be written by raid5d, so we don't
+			 * need to wait here.
+			 */
+			rdev_dec_pending(s.blocked_rdev,
+					 conf->mddev);
+	}
 
 	if (s.handle_bad_blocks)
 		for (i = disks; i--; ) {
@@ -3766,7 +3814,7 @@ static struct bio *remove_bio_from_retry(struct r5conf *conf)
 		 * this sets the active strip count to 1 and the processed
 		 * strip count to zero (upper 8 bits)
 		 */
-		bi->bi_phys_segments = 1; /* biased count of active stripes */
+		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
 	}
 
 	return bi;
@@ -3881,8 +3929,6 @@ static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
 		raid_bio->bi_next = (void*)rdev;
 		align_bi->bi_bdev =  rdev->bdev;
 		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);
-		/* No reshape active, so we can trust rdev->data_offset */
-		align_bi->bi_sector += rdev->data_offset;
 
 		if (!bio_fits_rdev(align_bi) ||
 		    is_badblock(rdev, align_bi->bi_sector, align_bi->bi_size>>9,
@@ -3893,6 +3939,9 @@ static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
 			return 0;
 		}
 
+		/* No reshape active, so we can trust rdev->data_offset */
+		align_bi->bi_sector += rdev->data_offset;
+
 		spin_lock_irq(&conf->device_lock);
 		wait_event_lock_irq(conf->wait_for_stripe,
 				    conf->quiesce == 0,
@@ -3962,6 +4011,62 @@ static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
 	return sh;
 }
 
+struct raid5_plug_cb {
+	struct blk_plug_cb	cb;
+	struct list_head	list;
+};
+
+static void raid5_unplug(struct blk_plug_cb *blk_cb, bool from_schedule)
+{
+	struct raid5_plug_cb *cb = container_of(
+		blk_cb, struct raid5_plug_cb, cb);
+	struct stripe_head *sh;
+	struct mddev *mddev = cb->cb.data;
+	struct r5conf *conf = mddev->private;
+
+	if (cb->list.next && !list_empty(&cb->list)) {
+		spin_lock_irq(&conf->device_lock);
+		while (!list_empty(&cb->list)) {
+			sh = list_first_entry(&cb->list, struct stripe_head, lru);
+			list_del_init(&sh->lru);
+			/*
+			 * avoid race release_stripe_plug() sees
+			 * STRIPE_ON_UNPLUG_LIST clear but the stripe
+			 * is still in our list
+			 */
+			smp_mb__before_clear_bit();
+			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
+			__release_stripe(conf, sh);
+		}
+		spin_unlock_irq(&conf->device_lock);
+	}
+	kfree(cb);
+}
+
+static void release_stripe_plug(struct mddev *mddev,
+				struct stripe_head *sh)
+{
+	struct blk_plug_cb *blk_cb = blk_check_plugged(
+		raid5_unplug, mddev,
+		sizeof(struct raid5_plug_cb));
+	struct raid5_plug_cb *cb;
+
+	if (!blk_cb) {
+		release_stripe(sh);
+		return;
+	}
+
+	cb = container_of(blk_cb, struct raid5_plug_cb, cb);
+
+	if (cb->list.next == NULL)
+		INIT_LIST_HEAD(&cb->list);
+
+	if (!test_and_set_bit(STRIPE_ON_UNPLUG_LIST, &sh->state))
+		list_add_tail(&sh->lru, &cb->list);
+	else
+		release_stripe(sh);
+}
+
 static void make_request(struct mddev *mddev, struct bio * bi)
 {
 	struct r5conf *conf = mddev->private;
@@ -3971,7 +4076,6 @@ static void make_request(struct mddev *mddev, struct bio * bi)
 	struct stripe_head *sh;
 	const int rw = bio_data_dir(bi);
 	int remaining;
-	int plugged;
 
 	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
 		md_flush_request(mddev, bi);
@@ -3990,7 +4094,6 @@ static void make_request(struct mddev *mddev, struct bio * bi)
 	bi->bi_next = NULL;
 	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
 
-	plugged = mddev_check_plugged(mddev);
 	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
 		DEFINE_WAIT(w);
 		int previous;
@@ -4089,24 +4192,19 @@ static void make_request(struct mddev *mddev, struct bio * bi)
 			finish_wait(&conf->wait_for_overlap, &w);
 			set_bit(STRIPE_HANDLE, &sh->state);
 			clear_bit(STRIPE_DELAYED, &sh->state);
-			if ((bi->bi_rw & REQ_SYNC) &&
+			if ((bi->bi_rw & REQ_NOIDLE) &&
 			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 				atomic_inc(&conf->preread_active_stripes);
-			release_stripe(sh);
+			release_stripe_plug(mddev, sh);
 		} else {
 			/* cannot get stripe for read-ahead, just give-up */
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
 			finish_wait(&conf->wait_for_overlap, &w);
 			break;
 		}
-			
 	}
-	if (!plugged)
-		md_wakeup_thread(mddev->thread);
 
-	spin_lock_irq(&conf->device_lock);
-	remaining = raid5_dec_bi_phys_segments(bi);
-	spin_unlock_irq(&conf->device_lock);
+	remaining = raid5_dec_bi_active_stripes(bi);
 	if (remaining == 0) {
 
 		if ( rw == WRITE )
@@ -4462,7 +4560,7 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 		     sector += STRIPE_SECTORS,
 		     scnt++) {
 
-		if (scnt < raid5_bi_hw_segments(raid_bio))
+		if (scnt < raid5_bi_processed_stripes(raid_bio))
 			/* already done this stripe */
 			continue;
 
@@ -4470,25 +4568,24 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 
 		if (!sh) {
 			/* failed to get a stripe - must wait */
-			raid5_set_bi_hw_segments(raid_bio, scnt);
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
 		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
 			release_stripe(sh);
-			raid5_set_bi_hw_segments(raid_bio, scnt);
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
+		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
 		handle_stripe(sh);
 		release_stripe(sh);
 		handled++;
 	}
-	spin_lock_irq(&conf->device_lock);
-	remaining = raid5_dec_bi_phys_segments(raid_bio);
-	spin_unlock_irq(&conf->device_lock);
+	remaining = raid5_dec_bi_active_stripes(raid_bio);
 	if (remaining == 0)
 		bio_endio(raid_bio, 0);
 	if (atomic_dec_and_test(&conf->active_aligned_reads))
@@ -4496,6 +4593,30 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 	return handled;
 }
 
+#define MAX_STRIPE_BATCH 8
+static int handle_active_stripes(struct r5conf *conf)
+{
+	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
+	int i, batch_size = 0;
+
+	while (batch_size < MAX_STRIPE_BATCH &&
+			(sh = __get_priority_stripe(conf)) != NULL)
+		batch[batch_size++] = sh;
+
+	if (batch_size == 0)
+		return batch_size;
+	spin_unlock_irq(&conf->device_lock);
+
+	for (i = 0; i < batch_size; i++)
+		handle_stripe(batch[i]);
+
+	cond_resched();
+
+	spin_lock_irq(&conf->device_lock);
+	for (i = 0; i < batch_size; i++)
+		__release_stripe(conf, batch[i]);
+	return batch_size;
+}
 
 /*
  * This is our raid5 kernel thread.
@@ -4506,7 +4627,6 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
  */
 static void raid5d(struct mddev *mddev)
 {
-	struct stripe_head *sh;
 	struct r5conf *conf = mddev->private;
 	int handled;
 	struct blk_plug plug;
@@ -4520,8 +4640,9 @@ static void raid5d(struct mddev *mddev)
 	spin_lock_irq(&conf->device_lock);
 	while (1) {
 		struct bio *bio;
+		int batch_size;
 
-		if (atomic_read(&mddev->plug_cnt) == 0 &&
+		if (
 		    !list_empty(&conf->bitmap_list)) {
 			/* Now is a good time to flush some bitmap updates */
 			conf->seq_flush++;
@@ -4531,8 +4652,7 @@ static void raid5d(struct mddev *mddev)
 			conf->seq_write = conf->seq_flush;
 			activate_bit_delay(conf);
 		}
-		if (atomic_read(&mddev->plug_cnt) == 0)
-			raid5_activate_delayed(conf);
+		raid5_activate_delayed(conf);
 
 		while ((bio = remove_bio_from_retry(conf))) {
 			int ok;
@@ -4544,21 +4664,16 @@ static void raid5d(struct mddev *mddev)
 			handled++;
 		}
 
-		sh = __get_priority_stripe(conf);
-
-		if (!sh)
+		batch_size = handle_active_stripes(conf);
+		if (!batch_size)
 			break;
-		spin_unlock_irq(&conf->device_lock);
-		
-		handled++;
-		handle_stripe(sh);
-		release_stripe(sh);
-		cond_resched();
+		handled += batch_size;
 
-		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
+		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
+			spin_unlock_irq(&conf->device_lock);
 			md_check_recovery(mddev);
-
-		spin_lock_irq(&conf->device_lock);
+			spin_lock_irq(&conf->device_lock);
+		}
 	}
 	pr_debug("%d stripes handled\n", handled);
 
@@ -4823,6 +4938,7 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	int raid_disk, memory, max_disks;
 	struct md_rdev *rdev;
 	struct disk_info *disk;
+	char pers_name[6];
 
 	if (mddev->new_level != 5
 	    && mddev->new_level != 4
@@ -4946,7 +5062,8 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
 		       mdname(mddev), memory);
 
-	conf->thread = md_register_thread(raid5d, mddev, NULL);
+	sprintf(pers_name, "raid%d", mddev->new_level);
+	conf->thread = md_register_thread(raid5d, mddev, pers_name);
 	if (!conf->thread) {
 		printk(KERN_ERR
 		       "md/raid:%s: couldn't allocate thread.\n",
@@ -5465,10 +5582,9 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	if (rdev->saved_raid_disk >= 0 &&
 	    rdev->saved_raid_disk >= first &&
 	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
-		disk = rdev->saved_raid_disk;
-	else
-		disk = first;
-	for ( ; disk <= last ; disk++) {
+		first = rdev->saved_raid_disk;
+
+	for (disk = first; disk <= last; disk++) {
 		p = conf->disks + disk;
 		if (p->rdev == NULL) {
 			clear_bit(In_sync, &rdev->flags);
@@ -5477,8 +5593,11 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 			if (rdev->saved_raid_disk != disk)
 				conf->fullsync = 1;
 			rcu_assign_pointer(p->rdev, rdev);
-			break;
+			goto out;
 		}
+	}
+	for (disk = first; disk <= last; disk++) {
+		p = conf->disks + disk;
 		if (test_bit(WantReplacement, &p->rdev->flags) &&
 		    p->replacement == NULL) {
 			clear_bit(In_sync, &rdev->flags);
@@ -5490,6 +5609,7 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 			break;
 		}
 	}
+out:
 	print_raid5_conf(conf);
 	return err;
 }
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 2164021f3b5..a9fc24901ed 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -210,6 +210,7 @@ struct stripe_head {
 	int			disks;		/* disks in stripe */
 	enum check_states	check_state;
 	enum reconstruct_states reconstruct_state;
+	spinlock_t		stripe_lock;
 	/**
 	 * struct stripe_operations
 	 * @target - STRIPE_OP_COMPUTE_BLK target
@@ -273,6 +274,7 @@ enum r5dev_flags {
 	R5_Wantwrite,
 	R5_Overlap,	/* There is a pending overlapping request
 			 * on this block */
+	R5_ReadNoMerge, /* prevent bio from merging in block-layer */
 	R5_ReadError,	/* seen a read error here recently */
 	R5_ReWrite,	/* have tried to over-write the readerror */
 
@@ -319,6 +321,7 @@ enum {
 	STRIPE_BIOFILL_RUN,
 	STRIPE_COMPUTE_RUN,
 	STRIPE_OPS_REQ_PENDING,
+	STRIPE_ON_UNPLUG_LIST,
 };
 
 /*